TWI309371B - Adaptive de-noising system and method - Google Patents

Description

1309371 VII. Designation of Representative Representatives: (1) The representative representative of the case is: (1). (2) The symbol of the symbol of this representative figure is simple: ίο Motion detection 20 Edge detection 30 Noise estimation 40 De-noise waver 8. If there is a chemical formula in this case, please reveal the chemical formula that best shows the characteristics of the invention: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive de-noising system and method, and more particularly to a method for predicting the generation of adaptive noise levels to perform different filtering operations. Information system and method. [Prior Art] When acquiring images via image components, noise is usually included, and these noises will affect the reality and effect of the image, and will affect the subsequent image processing quality. 1309371 There are some shortcomings in the prior art. For example, before predicting noise, it is necessary to predict the noise through a special standard system and a cumbersome calibration procedure. Another disadvantage is that a single independent signal source must be used to predict noise. SUMMARY OF THE INVENTION The present invention utilizes motion detection and edge detection to determine the motion and edge states of each element, thereby estimating a method for generating noise prediction, and using noise compensation operations in the time domain and the spatial domain to remove each 昼. The noise is used to improve the cumbersome calibration procedures of the prior art and the disadvantage of having to use a single independent signal source. One of the objects of the present invention is to use motion detection and edge detection to determine the motion and edge states of each element, and to estimate a noise level that is adaptively adjusted for the noise compensation operation of each element. Another object of the present invention is to use a spatial domain filter and a time domain filter to generate a denoising plane based on the moving flag, the amount of motion, the edge flag, and the variance of each pixel. [Embodiment] The following is a detailed description of the present invention, and the description of the adaptive denoising method and system in the following description does not include a detailed description of the detailed flow and operation principle. The prior art of the present invention is only used as a key reference here to assist the present invention.

Claims (1)

1309371, / t-day correction / ΐ ^ τιΐ ^ Γ 10, the scope of application for patents: 1. An adaptive de-noise system, including · · mobile detection device to detect the pixels in a face The edge detecting device is configured to detect an edge state of each element in the picture; the noise estimating device is used according to the motion detecting device, the motion detected by the edge detecting device, and the edge state. Estimating a modest noise level of the picture, wherein the noise level is fed back to the motion detection device, the edge detection device, and the noise removal filter device, according to the motion detection device, the edge detection The motion detection, the edge state, and the noise level detected by the measuring device perform a noise compensation operation of each pixel to obtain a de-noising surface. 2. The adaptive denoising system according to claim 1, wherein the motion detecting device is further configured to generate a difference between each pixel of the picture and a pixel corresponding to the adjacent pixel and the adjacent pixel. The weighted sum and the moving flag representing the moving state of each pixel. 3. The adaptive denoising system according to claim 2, wherein the edge detecting device is further configured to generate each pixel of the picture and a corresponding pixel between the adjacent pixel and the previous field. The average difference value and the edge flag representing the edge state of each pixel 0 17 1309371 f year, > month 'day correction / more responsibility 4. As described in claim 3, the adaptive de-noise system, wherein The above-mentioned noise estimation device is based on the average difference value between the moving flag representing the movement state of each element, the respective pixels and the neighboring pixels and the corresponding pixels of the previous field, and the edge state of each pixel. The edge flag is thus estimated to produce this adaptive noise level. 5. The adaptive denoising system according to claim 3, wherein the denoising filtering device is further based on a difference weighted sum of the pixels of the picture and the neighboring pixels and adjacent pictures. a moving flag representing the moving state of each pixel, an average difference between each pixel and its neighboring pixels and a corresponding pixel of the previous field, and an edge flag representing the edge state of each pixel to perform miscellaneous The compensation compensation is calculated. i 6. The adaptive denoising system according to claim 1, wherein the de-noising filtering device performs a spatial domain noise compensation operation and a time domain noise compensation operation. 7. The adaptive denoising system according to claim 3, wherein the moving flag representing the moving state of each pixel is a pixel between each pixel and its neighboring pixels. The difference weighted sum is generated as compared to a moving threshold. 18 1309371 A month '曰 Amendment/More 8. The adaptive de-noising system described in claim 7 of the patent application, wherein the above-mentioned edge flag representing the edge state of each element is a picture of each element and its neighbors. The average difference between the prime and the corresponding pixel of the previous field is compared with an edge threshold. 9. The adaptive de-noising system of claim 8, wherein the mobile threshold and the edge threshold are adaptively adjusted according to the size of the noise level. 10. An adaptive de-noising method, comprising: detecting a moving state of each element in a picture; i detecting an edge state of each pixel in the picture; according to the moving state and the edge state of the edge, Estimating an adaptive noise level of the picture, wherein the noise level feedback is used to adjust the moving threshold and the edge threshold; and performing each according to the moving state, the edge state, and the noise level The noise compensation operation of the pixel is obtained to obtain a noise removal surface. 11. The adaptive denoising method according to claim 10, wherein the moving state detecting step comprises: 19 1309371 playing the year ... month '曰 calculating each element of the face and its neighboring pixels and neighbors The weighted sum of the differences between the pixels corresponding to the pixels, so as to obtain the amount of elementary movement of each element of the picture; and determining whether the amount of movement of the element is greater than a moving threshold, thereby obtaining one of the elements of the element Move the flag. 12. The adaptive denoising method according to claim 10, wherein the edge state detecting step comprises: calculating between the pixels of the face and the neighboring pixels and the corresponding pixels of the previous field; The average difference value is used to obtain a variance number; and whether the variance number is greater than the edge threshold value, thereby obtaining an edge flag. I. 13. The adaptive noise removal method according to claim 10, wherein the adaptive noise level estimation step comprises: determining the pixel according to the movement state and the edge state in a plane Whether it is a static and smooth region; in a static and smooth region, accumulate the variance of each pixel; and normalize the accumulated variance to obtain the noise level. 14. The adaptive denoising method of claim 13, wherein the de-noising step comprises a spatial domain noise compensation operation and a time domain noise compensation operation. 20 1309371 15. The adaptive denoising method according to claim 14, wherein the spatial domain noise compensation operation step comprises: setting a pixel value of each pixel in a static and smooth region The average pixel value of adjacent pixels; in non-static and smooth regions, for non-marginal pixels, the pixel values of each element are set to the pixel values of each element according to the variation of each element. The weighted sum of the average pixel values of the neighboring pixels; and in the non-static and smooth regions, for the edge elements, the pixel values of the individual pixels are maintained. 16. The adaptive denoising method according to claim 14, wherein the time domain noise compensation operation step performs weighting calculation according to the spatial domain comparison result and the corresponding pixel value of the adjacent picture. —— 21 1309371 十一,图: f '-' 3⁄4 X,,· li 1 云尘 I ΙΑ 铖丨 rad^—fl»ll Local Motionlflag Msiosls JLo-'cavar Edige—flag Motionlflag Is 2 2ο ·55·ά”ίτ>,<61 in life B 22 1309371 23 1309371 Qing Yiyue Yiri Correction / 24 1309371 曰Correct, mmm,: a4 of >qS tal b穸varl=abs(bl*8-( b2+b3+a4+a5+a6+b7+b8+b9)); var2=aba(b2*8-( Bl+ b3+a4+a5+ae+b7+b8+b9)); mrnFMki (b3*8-( bl+b2f var 4=abs (a4*8- (bl+b2+b3+ ia5+a6+b7fb8+ii9) ); var5=abe(a5*8-(bl+b2+b3+a4+ a6+b7+b8+b9)); var6=al>e (a6«t-(bl+l>m34ii4^a5+, _Wei ν^Γ7=β^β(ΐτ7*®-(ΐ^^^3^4+β|τΝ||bS+M)>; vaf 8=abs (b8*8- (bl+b2+b3+ A4+a5+a6+b7+ b9)); var9=abs 〇>9*8- (bl+b2+b3+a4+a5+a6+b7+b8 )); lpcaIyar^sum=varlHhmr2^af34lraf|4vaf5^ Ai6^ar7#yar#+Tr»rSi l acal var=Loca i if (Iticalvar^dge-th) edge_f 1 agL«5sl \ else ejjg0jflag_j*5s:0i local third A crane 25 1309371 lying 3JSS 26 1309371 edgejQag If((lmotion ilag)&^(i〇〇〇〇Q total var sum+^localvar, rnmt- emm. · total_var_coimt-H-; Local· Fourth A reading 27 1309371 Factory year 11 k fiB surface 28 1309371 6曰Revised The first straight A8 c8 Ιοί mm α7 α8 α9 nmsejflag loealltoF (Βΐ+ΐΜ)_13 Wei 1 face 5+^89)/8; xf job LEVmi«»lVar]; y=MOTON_LEVEL[m^0n_stt]; efejElag ^ motioii one Flag ' ifi[noise-flag) a8'=localMcan; else into ledge-flag) β8·=(χ*α8-Κ16- x)*localMemi)/16; else aS^aS; localVar 'motion one sun > aSHy *aSH<16^fc Lai; Fifth Boer 29 1309371 30