WO2014112076A1 - Video encoding device - Google Patents

Abstract

An encoding unit (10) encodes difference values between a frame to be encoded and a predicted frame. A scene change detection unit (12) detects the presence or absence of scene changes. A luminance block accumulator (13) calculates the difference for each cumulative value of the luminance of corresponding blocks between the current picture and the first previous referable picture, denoting the cumulative value for one picture as X. A luminance accumulator (14) calculates the difference between each of: the difference (Y1) between cumulative values of the luminance between the first previous referable picture and the current picture; the difference (Y2) between cumulative values of the luminance of the second previous referable picture and the current picture; and the difference (Y3) between cumulative values of luminance between the third previous referable picture and the current picture. A flash assessment unit (15) assesses a flash image on the basis of the differences between a scene change detection pattern, X, and Y1 to Y3. A predicted frame calculation unit (9) calculates a predicted frame by referencing the referable picture just before the flash image.

Description

Video encoding device

The present invention relates to a moving picture coding apparatus that performs coding using inter-frame prediction, and more particularly to a moving picture coding apparatus that can prevent image disturbance immediately after flashing.

Inter-frame prediction is used in video compression. However, since a flash image has a low degree of similarity with an image adjacent thereto, it is often inappropriate to use the flash image as a reference frame in inter-frame prediction. Therefore, it has been proposed to use the frame before the flash as the reference frame of the frame after the flash (see, for example, Patent Document 1).

JP 2007-166408 A

In the prior art, a flash image is detected by using two of the inter-frame difference value of the accumulated luminance value and the difference for each pixel of the line in the first direction / first direction. However, since the difference for each pixel of the line in the first direction / first direction is too sensitive to movement, the difference increases only by moving one pixel from the previous frame, and false detection increases. Even when only the inter-frame difference value of the accumulated luminance value is used, erroneous detection increases. As a result, there is a problem that the image immediately after the flash is disturbed.

The present invention has been made to solve the above-described problems, and an object of the present invention is to obtain a moving picture coding apparatus capable of preventing image disturbance immediately after flashing.

The present invention is a moving image encoding apparatus that performs encoding using inter-frame prediction, and includes an encoding unit that encodes a difference value between an encoding target frame and a prediction frame, and a current picture of the encoding target frame , A scene change detection unit for detecting the presence or absence of a scene change for each of the previous and second previous referenceable pictures, and the luminance of a block corresponding between the current picture and the previous referenceable picture A luminance block accumulating unit that calculates X as the difference of each accumulated value accumulated by one picture, and the difference between the accumulated luminance values of the previous referenceable picture and the current picture as Y1 , The difference between the accumulated values of the luminances of the previous referenceable picture and the current picture is Y2, and the previous referenceable picture and the current picture are (1) The current picture and the 1 are the difference between the accumulated luminance values of Y3, Y3, the luminance accumulation unit for calculating the difference between them, the first threshold value Xth, and the second threshold value Yth. If a scene change is detected in the previous referenceable picture and | X | ≧ Xth and | Y1-Y2 | ≧ Yth, the previous referenceable picture is determined to be a flash image, and (2) the one When a scene change is detected between the previous referenceable picture and the previous two referenceable pictures, and | X | ≧ Xth and | Y2-Y3 | ≧ Yth, the previous and second previous referenceable pictures are (3) When a scene change is detected in the current picture and the previous referenceable picture, and | X | ≧ Xth and | Y1-Y3 | ≧ Yth, the previous and second Previous It is determined that the illuminable picture is a flash image. (4) If a scene change is detected in the current picture and the previous referenceable picture, and | X | ≧ Xth and | Y1-Y3 | ≧ Yth, A flash determination unit that determines that the previous and second previous reference pictures are flash images, and refers to the first previous reference picture when the first previous reference picture is not determined to be a flash image; The prediction frame is calculated, and when the previous referenceable picture is determined to be a flash image, the prediction frame is calculated with reference to the two previous referenceable pictures, and the previous and second Prediction frame calculation for calculating the prediction frame with reference to the three previous referenceable pictures when the previous referenceable picture is determined to be a flash image And a section. Here, the block size may be considered arbitrary. The larger the block size, the more moderate the increase / decrease of the difference value with respect to the motion.

According to the present invention, it is possible to prevent image disturbance immediately after flashing.

It is a block diagram which shows the moving image encoder which concerns on embodiment of this invention. It is a figure which shows the relationship between a flash image and a scene change. It is a figure which shows the calculation method of a brightness | luminance block accumulation part. It is a figure which shows the calculation method of a brightness | luminance accumulation part. It is a figure which shows the example of selection of the picture referred when calculating a prediction frame. It is a figure which shows the example of selection of the picture referred when calculating a prediction frame.

A moving picture encoding apparatus according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a moving picture coding apparatus according to an embodiment of the present invention. This moving image encoding apparatus performs encoding using inter-frame prediction.

The input moving image signal is a moving image signal (video signal) reproduced by a video recording / reproducing device such as a digital VTR or a DVD system capable of repeatedly reproducing the same signal a plurality of times. There are three types of image types of I picture, P picture, and B picture in the encoding target frame of the input video signal.

An I picture is a picture that is encoded only with its own picture information and is the most beautiful but requires the most code amount. A P picture is a picture that is encoded only with its own picture information and is encoded with reference to a picture earlier than itself, and is more beautiful than a B picture, but has a larger amount of code than a B picture. A B picture is a picture that is encoded with reference to a past / future picture from itself, and the amount of code can be reduced most among three pictures. In general, an I picture and a P picture are referenceable pictures that can be referred to when other pictures are encoded, while a B picture is a non-referenceable picture.

Video interface 1 inputs an input video signal and outputs a frame to be encoded. The subtracter 2 calculates the difference between the encoding target frame and the predicted frame. The discrete cosine transformer 3 performs a discrete cosine transform (DCT) on the output of the subtracter 2 in units of a fixed block. The quantization unit 4 quantizes the DCT coefficient data obtained by the discrete cosine transform.

The inverse quantization unit 5 performs inverse quantization on the quantized DCT coefficient data. The inverse discrete cosine transformer 6 performs inverse discrete cosine transform (inverse DCT) on the inversely quantized DCT coefficient data. The adder 7 adds the output of the inverse discrete cosine transformer 6 and the prediction frame, and stores the addition result in the video memory 8 as a referenceable picture.

The motion compensation prediction unit 9 does not perform motion vector detection when the encoding target frame is an I picture, and outputs 0 as a prediction frame. When the encoding target frame is a P picture or a B picture, the motion compensation prediction unit 9 starts from the encoding target frame. A motion vector is detected for each macroblock, and a prediction frame is generated.

Motion vector detection is performed in units of macroblocks between the encoding target frame and the referenceable picture stored in the video memory 8. Specifically, by detecting the macroblock having the highest correlation of the referenceable picture with respect to the macroblock of the encoding target frame, the macroblock of the encoding target frame has moved from which macroblock of the referenceable picture. Is detected as a motion vector. The motion compensation prediction unit 9 performs motion compensation on the referenceable picture using this motion vector to calculate a prediction frame.

The motion compensation prediction unit 9 also outputs a prediction mode of motion compensation prediction and prediction mode / motion vector information indicating a motion vector. The entropy encoding unit 10 encodes the quantized DCT coefficient data and the prediction mode / motion vector information. In this way, the entropy encoding unit 10 encodes the difference value between the encoding target frame and the prediction frame.

The header / syntax adding unit 11 adds syntax information to the encoded result. A number ref-idx is assigned as syntax information to an image to be a predicted frame. Ref-idx is assigned in order from 0 in order of time closest to the current picture being processed.

The scene change detection unit 12 detects the presence / absence of a scene change (SC) for each of the current picture of the encoding target frame, the previous one picture, and the second previous reference picture. FIG. 2 is a diagram showing the relationship between a flash image and a scene change. When the entire surface is a flash image only one picture before, a scene change is detected between the current picture and the previous referenceable picture (pattern 1). When the entire front surface of 2 pictures is a flash image, and 1/3 of the screen before 1 picture is a flash image (1/3 of the screen is white), the previous referenceable picture and the previous 2 pictures A scene change is detected in the referenceable picture (pattern 2). If 1/3 of the screen two pictures before is a flash image and the entire picture one picture before is a flash image, a scene change is detected between the current picture and the previous referenceable picture (pattern 3). If the entire front surface of two pictures is a flash image and 2/3 of the screen before the one picture is a flash image (2/3 of the screen is white), the current picture and the previous referenceable picture A scene change is detected (pattern 4). If 2/3 of the screen two pictures before is a flash image and the entire face before one picture is a flash image, a scene change is detected between the current picture and the two previous referenceable pictures (pattern 5).

FIG. 3 is a diagram showing a calculation method of the luminance block accumulation unit. The luminance block accumulating unit 13 calculates, as X, a value obtained by accumulating the difference for each accumulated value of luminance of the corresponding block between the current picture and the previous referenceable picture for one picture. The block is a macroblock of 16 pixels × 16 pixels here, but is not limited thereto, and may be a block of 4 pixels × 4 pixels, 8 pixels × 8 pixels, 16 pixels × 32 pixels, or the like.

FIG. 4 is a diagram illustrating a calculation method of the luminance accumulation unit. The luminance accumulation unit 14 sets Y1 as the difference between the accumulated luminance values of the previous referenceable picture and the current picture, and determines the difference between the accumulated luminance values of the previous referenceable picture and the current picture. Y2 is set as Y2, and the difference between the accumulated values of the luminance of the three previous referenceable pictures and the current picture is set as Y3, and the difference between them is calculated.

The flash determination unit 15 determines the flash image as follows by setting the first threshold value to Xth and the second threshold value to Yth.
(1) When a scene change is detected between the current picture and the previous referenceable picture, and | X | ≧ Xth and | Y1-Y2 | ≧ Yth, the previous referenceable picture is determined to be a flash image ( Corresponds to pattern 1 in FIG.
(2) When a scene change is detected in the previous referenceable picture and the previous referenceable picture, and | X | ≧ Xth and | Y2-Y3 | ≧ Yth, the previous and second previous references A possible picture is determined as a flash image (corresponding to pattern 2 in FIG. 2).
(3) When a scene change is detected between the current picture and the previous referenceable picture, and | X | ≧ Xth and | Y1-Y3 | ≧ Yth, the previous and second previous referenceable pictures are flash images. (Corresponding to pattern 3 in FIG. 2).
(4) When a scene change is detected between the current picture and the previous referenceable picture, and | X | ≧ Xth and | Y1-Y3 | ≧ Yth, the previous and second previous referenceable pictures are flash images. (Corresponding to patterns 4 and 5 in FIG. 2).

Then, when the flash determination unit 15 detects the flash image, the flash determination unit 15 instructs to change the predicted frame to the frame immediately before the flash image. Upon receiving this instruction, the motion compensated prediction unit 9 calculates a prediction frame by referring to the previous referenceable picture when the previous referenceable picture is not determined to be a flash image, and calculates the previous frame. When the referenceable picture is determined to be a flash image, a prediction frame is calculated with reference to the previous referenceable picture, and when the previous and second previous referenceable pictures are determined to be flash images, 3 is calculated. A prediction frame is calculated with reference to the previous referenceable picture.

5 and 6 are diagrams showing examples of selecting pictures to be referred to when calculating a prediction frame. FIG. 5 shows a case where the interval M to the referenceable picture (I or P picture) is 1. In the case of 1 picture flash, the previous picture is referred to, and in the case of 2 picture flash, the previous picture is referred to so that the flash image is not referred to.

FIG. 6 shows a case where M is 3, the interval N to the next I picture is 15, and the fourth P picture is a flash image. When the 7th P picture is encoded, the 1st I picture is referred to without referring to the 4th P picture.

If the I picture is a flash picture, the picture immediately after that is changed to a P picture and encoded. After that, encoding is performed with a GOP configuration adapted to the change. Although the immediately following picture may be an I picture, a large amount of codes are generated. If the B picture is a flash image, the B picture is not referred to in the first place, so that encoding is performed as usual.

Also, since the code amount is reduced as the ref-idx of the referenced image is smaller, it is a waste to assign the number ref-idx = 0 that gives the most advantageous code amount to the flash image that is not referenced. Therefore, when the flash determination unit 15 detects the flash image, the flash determination unit 15 instructs the header / syntax adding unit 11 to change ref-idx of the frame immediately before the flash image to 0.

In the present embodiment, the calculation result X of the luminance block accumulating unit 13 is more responsive to the motion of the picture (increase / decrease in the difference value) than the accumulated luminance difference for each pixel. The difference between Y1, Y2, and Y3 calculated by the luminance accumulation unit 14 is more responsive to the motion of the picture (increase / decrease in the difference value) than the calculation result X of the luminance block accumulation unit 13. Accordingly, since the flash image is determined for each scene change calculation pattern based on these calculation results, erroneous detection of the flash image is reduced. Further, it is possible to cope with a two-picture flash in which a flash is generated over two frames. For this reason, the disturbance of the image immediately after the flash can be prevented.

9 motion compensated prediction unit (predicted frame calculation unit), 10 entropy coding unit (coding unit), 12 scene change detection unit, 13 luminance block accumulation unit, 14 luminance accumulation unit, 15 flash determination unit

Claims (1)

A video encoding device that performs encoding using inter-frame prediction,
An encoding unit that encodes a difference value between the encoding target frame and the prediction frame;
A scene change detection unit for detecting the presence or absence of a scene change for the current picture of the encoding target frame, the previous and second referenceable pictures,
A luminance block accumulating unit that calculates, as X, a difference obtained by accumulating the luminance value of the corresponding block between the current picture and the previous referenceable picture for one picture;
The difference between the accumulated luminance values of the previous referenceable picture and the current picture is Y1, and the difference between the accumulated luminance values of the previous referenceable picture and the current picture is Y2. A luminance accumulation unit for calculating a difference between Y3 as a difference between the luminance accumulation values of the three previous referenceable pictures and the current picture;
When the first threshold value is Xth and the second threshold value is Yth, (1) a scene change is detected in the current picture and the previous referenceable picture, and | X | ≧ Xth and | Y1-Y2 | ≧ In the case of Yth, the previous referenceable picture is determined as a flash image, and (2) a scene change is detected in the previous referenceable picture and the second previous referenceable picture, and | X | ≧ If Xth and | Y2-Y3 | ≧ Yth, the previous and second previous referenceable pictures are determined to be flash images, and (3) a scene change occurs between the current picture and the previous previous referenceable picture. If | X | ≧ Xth and | Y1-Y3 | ≧ Yth, the previous and second previous referenceable pictures are determined to be flash images, and (4) the current picture and the previous When a scene change is detected in the previous referenceable picture, and | X | ≧ Xth and | Y1−Y3 | ≧ Yth, the flash determination for determining the previous and second previous referenceable pictures as flash images And
When the previous referenceable picture is not determined to be a flash picture, the prediction frame is calculated with reference to the previous referenceable picture, and the previous referenceable picture is determined to be a flash picture. The prediction frame is calculated by referring to the previous two referenceable pictures, and the third previous reference is determined when the previous and second previous referenceable pictures are determined to be flash images. A motion picture encoding apparatus comprising: a prediction frame calculation unit that calculates the prediction frame with reference to a possible picture.

Images