201027982 九、發明說明： 【發明所屬之技術領域】 本發明係關於一種影像處理裝置’特別是關於—種可 防止操作時因手震動而造成機器晃動使得影像模糊之影 像處理裝置。 【先前技術】 一 隨著手持式攝影裝置的普及，一般人在日常生活中常 ® 常有機會使用這些裝置來紀錄影像，令人困擾的是一般人 並沒受過專業的攝影訓練，常常在需要較長時間的拍攝時 因為手的晃動造成拍攝影像的失焦、模糊，使得所拍攝影 像品質不佳。 習知技術中有許多方式來改善此現象，大致可分為機 械式補償以及電子式補償，後者大多是比較兩拍攝影像間 的差異或者是某塊特定影像區域與週遭背景的變化來判 斷兩影像間是不是有因手震而造成的晃動，其所比較的對 〇 象是實際物體經感光元件擷取影像後經由影像處理元件 ： 處理過的影像，換句話說，其所比對的影像是實際輸出至 • 顯不器（例如LCD)之影像，或者說是已經轉換成γυν域 (YUV domain)之影像’此種影像資料非常精確詳細，但 是也因此而資料量非常龐大’以Full HD規格來說，至少 就需1920χ1〇80χ2χ2位元組的資料量來做兩影像間是否有 移動的判斷，使得必須採用頻寬與容量皆較大的記憶體。 因此，如何提供一種判斷手震以及補償該手震造成影 6 201027982 響的影像處理裝置使得所需處理的資料量減少，是本發明 主要的目的，所需處理的資料量愈少，處理的速度便愈 快，所需的記憶體也愈少，對手持式攝影裝置的成本、尺 寸等都有正面助益。 ： 【發明内容】 針對上述問題，本發明之目的在提供一種可防止操作 時因手晃動而造成機器晃動使得影像模糊之影像處理裝 ^ 置與影像處理方法。 為達成上述目的，本發明之一種防手震影像處理裝 置，包含.一感測單元，用來產生一貝爾圖樣影像（Bayer Pattern image); 一取樣轉換單元，耦接於該影像感測單 元，用來取樣該貝爾圖樣影像中一原色並將該原色轉換為 一原色資料；一移動判斷單元，耦接於該取樣轉換單元， 用來比較前後兩圖框（frame )之該原色資料並產生一移動 向量；以及一影像處理單元，耦接於該移動判斷單元與該 G 感測單元，用來根據該移動向量修正一影像並輸出該修正 : 後之影像。 - 本發明另提出一種可防手震之影像處理方法，包含： 產生一貝爾圖樣影像（Bayer Pat tern image );取樣該貝 爾圖樣影像中一原色並將該原色轉換為一原色資料；比較 前後兩圖框（frame)之該原色資料並產生一移動向量；以 及根據該移動向量修正一影像並輸出該修正後之影像。 藉由直接在影像處理流程前端的貝爾圖樣影像中取 7 201027982 樣三原色中的-種原色資料作為偵測連續兩圖框中之影 像是否移動的依據’移動判斷單元所需處理的資料量大大 減少，以Full HD規格來說，移動判斷單元只需72〇 μ⑽以 位元組的資料量便可執行移㈣測，㈣只需咖倾規 格記憶體頻寬即可運作，而習知技術則必須使用較昂貴之 DDRRAM記憶體。 【實施方式】 ⑩ 冑參考第1圖，第1圖為本發明防手震之影像處理裝 置10 —實施例方塊圖，如圖所示，本實施例包含有一感 測單元20、一取樣轉換單元30、一移動判斷單元切以及 一影像處理單元50。感測單元20包含有CCD元件及濾光 片等元件，擷取外界影像後形成如第2A圖所示之貝爾圖 樣影像（Bayer Pattern image )，亦即將擷取到的外界影像 分解成若干個像素，而每一像素包含有紅 '綠、藍（R(jB) 三種原色資料，此原色資料經取樣轉換單元3〇轉換後， 〇 其中一種原色被取樣並經一轉換過程成為一原色資料，如 - 同第2B圖所示。請注意，原則上取樣紅、綠、藍三種原 : 色任一種皆可達到本發明之目的，但是由於肉眼對綠色的 感文最強烈’綠色的亮度對比分佈也最符合實際上真實影 像的亮度對比，所以本實施例以取樣綠色為例，但實施範 圍卻不以綠色為限。接下來，此像素與轉換後綠原色亮度 的對應資料會被傳送至移動判斷單元40，移動判斷單元 40會將前一個圖框的原色資料先儲存下來，等到接收到下 8 201027982 一個圖框的原色資料時，移動判斷單元40比對這兩個連 續圖框中相對應像素的原色資料以得出一移動向量，例如 一圖框之某區塊中像素的原色資料值皆與前一圖框該區 • 塊左邊另$塊中像素的原色資料值相同，則可推論在此 - 」圖框取像時間中攝影鏡頭向右晃動，此種移動向量的估 : 算方式有很多種，在此不再贅述。影像處理單元5〇主要 的功能是將原先擷取到的貝爾圖樣影像還原成一般肉眼BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus, and more particularly to an image processing apparatus which can prevent image blurring caused by hand shake due to hand shake during operation. [Prior Art] With the popularity of handheld photographic devices, most people in daily life often have the opportunity to use these devices to record images. It is disturbing that most people do not have professional photography training, and often take a long time. The image was out of focus and blurred due to the shaking of the hand during shooting, which made the captured image quality poor. There are many ways to improve this phenomenon in the prior art, which can be roughly divided into mechanical compensation and electronic compensation. The latter mostly compares the difference between two captured images or the change of a specific image area and the surrounding background to judge the two images. Is there any sway caused by the shaking of the hand, the opposite object is that the actual object is captured by the photosensitive element and then passes through the image processing component: the processed image, in other words, the compared image is The actual output to the image of the display (such as LCD), or the image that has been converted to the γυν domain (YUV domain). This image data is very accurate and detailed, but the data volume is very large. In other words, at least the amount of data of 1920χ1〇80χ2χ2 bytes is required to determine whether there is movement between the two images, so that memory with a large bandwidth and capacity must be used. Therefore, how to provide a kind of image processing device for judging the hand shake and compensating for the hand shake to reduce the amount of data to be processed is the main purpose of the present invention, and the less the amount of data to be processed, the processing speed The faster the memory, the less memory is required, which is positive for the cost and size of the handheld camera. SUMMARY OF THE INVENTION In view of the above problems, an object of the present invention is to provide an image processing apparatus and an image processing method which can prevent image blurring caused by hand shake due to hand shake during operation. In order to achieve the above object, an anti-shake image processing apparatus of the present invention includes: a sensing unit for generating a Bayer Pattern image; a sampling conversion unit coupled to the image sensing unit, For sampling a primary color in the Bell pattern image and converting the primary color into a primary color data; a movement determining unit coupled to the sampling conversion unit for comparing the primary color data of the front and rear frames (frame) and generating a And a motion processing unit coupled to the motion determining unit and the G sensing unit, configured to correct an image according to the motion vector and output the corrected image: - The invention further provides an image processing method capable of preventing hand shake, comprising: generating a Bayer Pat tern image; sampling a primary color in the Bell pattern image and converting the primary color into a primary color data; Forming the primary color data of the frame and generating a motion vector; and correcting an image according to the motion vector and outputting the corrected image. By directly taking the primary color data in the three primary colors of the 201027982 sample in the Bell pattern image at the front end of the image processing flow as the basis for detecting whether the images in the two consecutive frames are moved, the amount of data required for the movement determination unit is greatly reduced. In the case of Full HD, the mobile judgment unit only needs 72〇μ(10) to perform shift (four) measurement with the data amount of the byte, and (4) it can operate only by the specification of the memory bandwidth, while the conventional technology must Use more expensive DDRRAM memory. [Embodiment] 10 胄 Referring to FIG. 1 , FIG. 1 is a block diagram of an image processing apparatus 10 for anti-shake detection of the present invention. As shown in the figure, the present embodiment includes a sensing unit 20 and a sampling conversion unit. 30. A movement judging unit cuts and an image processing unit 50. The sensing unit 20 includes components such as a CCD element and a filter, and extracts an external image to form a Bayer Pattern image as shown in FIG. 2A, and also decomposes the captured external image into a plurality of pixels. Each pixel contains red 'green, blue (R (jB) three primary color data, after the primary color data is converted by the sampling conversion unit 3, one of the primary colors is sampled and converted into a primary color data through a conversion process, such as - Same as shown in Figure 2B. Please note that in principle, the three original colors of red, green and blue are sampled: any one of the colors can achieve the purpose of the present invention, but since the naked eye has the strongest sense of green, the green brightness contrast distribution is also It is the most suitable for the brightness comparison of the actual image. Therefore, the sampling green is taken as an example, but the implementation range is not limited to green. Next, the corresponding data of the pixel and the converted green primary color brightness will be transmitted to the mobile judgment. The unit 40, the movement judging unit 40 stores the original color data of the previous frame first, and waits until the primary color data of a frame of the next 8 201027982 is received, and the movement judgment sheet is moved. 40 compares the primary color data of the corresponding pixels in the two consecutive frames to obtain a motion vector. For example, the primary color data values of the pixels in a certain block of a frame are the same as the previous frame. If the values of the primary color data of the pixels in the block are the same, it can be inferred that the photographic lens is swaying to the right during the image capturing time. The estimation of the moving vector has many calculation methods, and will not be described here. The main function of the 5 是 is to restore the original captured Bell pattern image to the general naked eye.
φ 熟悉的現實影像，包括了將原先臟域信號轉換成YUV 域之影像信號、將影像像素值放大成後端顯示器所須之像 素值、色彩調整、影像亮度、對比度、色彩飽和度等等的 影像處理，在決定輸出影像前影像處理單元50會根據移 動判斷單元40估算出的移動向量修正最後輸出的影像， 例如移動向量顯示鏡頭向下晃動，則影像處理單元5 〇在 操取該圖框影像時會向上修正，使得該晃動造成的影響不 會顯示在輸出影像上。 第1圖中取樣轉換單元3〇決定取樣貝爾圖樣影像中 ® ㈣原色後會將綠原色的亮度值作轉換，將各像素的亮度 - 對比值拉大以使得後端的影像處理與移動向量估算更容 易進行„青參考第3圖，第3圖為g⑽咖轉換時轉換前 像^冗度值與轉換後像素亮度值的函數曲線圖，橫軸是轉 換别的亮度灰階，以8位元的灰階值來說共分為從〇至 255的256階；縱轴則是轉換後的亮度灰階，一樣分為2% 階，請參考第4A圖，帛4A圖為經取樣轉換單元％作 Gamma轉換前的像素綠原色亮度直方圖（histog職），橫 9 201027982φ Familiar reality images, including image signals that convert the original visceral domain signal into the YUV domain, magnify the image pixel values into the pixel values required for the back-end display, color adjustment, image brightness, contrast, color saturation, etc. For image processing, the image processing unit 50 corrects the last output image according to the motion vector estimated by the movement determining unit 40 before determining the output image. For example, if the motion vector display lens is shaken downward, the image processing unit 5 is operating the frame. The image is corrected upwards so that the effects of the sway are not displayed on the output image. In the first figure, the sampling conversion unit 3 determines the color of the green primary color after converting the (4) primary color in the sample of the Bell pattern, and enlarges the brightness-contrast value of each pixel to make the image processing and motion vector estimation of the back end more. It is easy to carry out the blue reference picture 3, the third picture shows the function curve of the image before the conversion and the brightness value of the converted pixel when the g(10) coffee conversion is performed, and the horizontal axis is the conversion of other brightness gray scales to 8 bits. The gray scale value is divided into 256 steps from 〇 to 255; the vertical axis is the converted gray level, which is divided into 2% steps, please refer to Figure 4A, 帛4A is the sample conversion unit % Pixel green primary color brightness histogram before Gamma conversion (histog job), horizontal 9 201027982
軸表不256階的灰階值，縱轴表示一圖框中具有該灰階亮 度的像素數目，由帛4A圖可知該圖框中大部分像素亮度 值都偏暗，也就是說各像素間亮度差異值不大，這對於之 後比對兩連續圖框以估算位移向量的步驟帶來不便，也不 利於後續輸出影像的對比度，因此取樣轉換單& %將原 來第4A圖所示圖框的像素綠原色亮度值經過第3圖的 Gamma轉換後，便成為如第4B圖所示之像素綠原色亮度 直方圖，很明顯的，轉換後各像素的亮度值均勻的分布在 各灰階值中。請注意’本實施例雖以Gamma轉換為例解 釋取樣轉換單元3〇的轉換動作，但實際上若不以Gamma 函數做轉換，而另取一函數將原始像素綠原色亮度轉換成 一較均勻分布的原色資料亦屬本發明揭露之範圍。 請參考第5圖，第5圖為第1圖中移動判斷單元4〇 的方塊圖，移動判斷單元40包含一組圖框緩衝器（frame buffer )410、一移動判斷器420、一移動向量產生器43〇 以及一移動向量緩衝器440，本實施例以三個圖框緩衝器 組成該組圖框緩衝器410方便說明，但實際上圖框緩衝器 數目可視設計時之需求而改變，移動向量產生器43〇内含 一移動向量表（motion vector table ) 43 1與一影像穩定分 析器（video stable analyzer ) 432。取樣轉換單元3〇將兩 連續圖框（設為第一圖框與第二圖框）的原色資料輸入至 移動判斷單元40後，資料會先被儲存在第一與第二圖框 緩衝器，接著移動判斷器420將第一圖框與第二圖框的原 色資料自圖框緩衝器中讀出並作比較，同時第三圖框的原 10 201027982 色資料被輸入並儲存在第三圖框緩衝器，移動判斷器42〇 將比較後產生的比較值輸入移動向量產生器43〇，並從第 三圖框緩衝器讀出第三圖框的原色資料並與第二圖框的 原色資料作比較，此時第四圖框的原色資料被輸入並儲存 • 在第一圖框緩衝器，移動判斷器420將比較第二圖框與第 . 二圖框的原色資料後產生的比較值輸入移動向量產生器 430後繼續從第一圖框緩衝器讀出第四圖框的原色資料並 φ 與第二圖框的原色資料作比較，同時第五圖框的原色資料 則被輸入並儲存在第二圖框緩衝器，如此重複上述動作， 則一連串的連續圖框原色資料被輸入並作比較得出一連 串比較值被輸入至移動向量產生器43〇而不會中斷。移動 向量產生器430内含一移動向量表43丨，可由查表得知不 同比較值所對應的移動向量大小與方向，此初始移動向量 經由影像穩定分析器432調整並得出一圖框之總合移動向 量並輸入移動向量緩衝器440，最後，影像處理單元5〇要 輸出°亥圖框影像時會從移動向量缓衝器440讀出該圖框之 和移動向量並藉此修正最後輸出之影像。 凊參考第6圖，第6圖為本發明實施例之詳細方塊 • 圖，其中感測單元20、取樣轉換單元3〇、移動判斷單元 、及八刀別所包含之功能元件已於前述’而影像處理單 " 貝】匕含有一影像仏號處理器510、一縮放器（scaler ) 520、—輸出緩衝器53〇、一輸出緩衝控制器54〇以及一輸 出"面控制器550。取樣轉換單元30將貝爾圖樣資料輸入 至影像處理單元50後影像信號處理器510會將此貝爾圖 11 201027982The axis table does not have a gray scale value of 256 steps, and the vertical axis represents the number of pixels having the gray scale brightness in a frame. As shown in FIG. 4A, most of the pixel brightness values in the frame are dark, that is, between pixels. The brightness difference value is not large, which is inconvenient for the step of comparing the two consecutive frames to estimate the displacement vector, and is not conducive to the contrast of the subsequent output image, so the sample conversion list & % will be the original frame shown in FIG. 4A. The luminance value of the pixel green primary color is converted into the histogram of the green primary color of the pixel as shown in FIG. 4B after the gamma conversion of FIG. 3, and it is obvious that the luminance value of each pixel after the conversion is uniformly distributed in each grayscale value. in. Please note that the present embodiment uses the gamma conversion as an example to explain the conversion operation of the sampling conversion unit 3 ,, but in fact, if the conversion is not performed by the Gamma function, another function is used to convert the original pixel green primary color brightness into a more evenly distributed one. Primary color data is also within the scope of the present invention. Please refer to FIG. 5, which is a block diagram of the movement determining unit 4A in FIG. 1. The movement determining unit 40 includes a set of frame buffer 410, a motion determiner 420, and a motion vector generation. The device 43A and a motion vector buffer 440, in this embodiment, the frame buffer 410 is composed of three frame buffers for convenience, but in fact, the number of frame buffers can be changed according to the design requirements, and the motion vector is changed. The generator 43A includes a motion vector table 43 1 and a video stable analyzer 432. The sample conversion unit 3 inputs the primary color data of the two consecutive frames (set as the first frame and the second frame) to the movement determination unit 40, and the data is first stored in the first and second frame buffers. Then, the movement determiner 420 reads and compares the primary color data of the first frame and the second frame from the frame buffer, and the original 10 201027982 color data of the third frame is input and stored in the third frame. The buffer, the motion determiner 42 inputs the comparison value generated after the comparison into the motion vector generator 43A, and reads the primary color data of the third frame from the third frame buffer and performs the primary color data of the second frame. In comparison, the primary color data of the fourth frame is input and stored at this time. • In the first frame buffer, the movement determiner 420 compares the comparison value generated after comparing the second frame and the primary color data of the second frame. After the vector generator 430 continues to read the primary color data of the fourth frame from the first frame buffer and φ is compared with the primary color data of the second frame, and the primary color data of the fifth frame is input and stored in the first Two frame buffers, so heavy The above operation, the series of continuous primary color frame information is inputted and compared consecutive draw string comparison value is input to the motion vector generator 43〇 without interruption. The motion vector generator 430 includes a motion vector table 43丨, and the size and direction of the motion vector corresponding to the different comparison values can be obtained by looking up the table. The initial motion vector is adjusted by the image stabilization analyzer 432 to obtain a total frame. The motion vector is combined and input into the motion vector buffer 440. Finally, when the image processing unit 5 outputs the image, the motion vector buffer 440 reads the sum vector of the frame and corrects the final output. image. Referring to FIG. 6, FIG. 6 is a detailed block diagram of an embodiment of the present invention, wherein the sensing unit 20, the sampling conversion unit 3, the movement judging unit, and the functional components included in the eight-knife have been in the foregoing image. The processing list contains an image nickname processor 510, a scaler 520, an output buffer 53A, an output buffer controller 54A, and an output "face controller 550. The sample conversion unit 30 inputs the Bell pattern data to the image processing unit 50, and the image signal processor 510 will present the Bell diagram 11 201027982
樣影像轉換成一般顯示裝置能辨識的YUV域影像信號， 一般來說’ YUV域影像信號的資料量會遠大於貝爾圖樣影 像，而影像信號處理器510也另外會執行一些與色彩、飽 和度、對比度等影像效果。由於後端所接的顯示裝置不 同’此YUV域影像信號所需滿足的顯示解析度也不同， 以Full HD規格來說，便需滿足192〇χ1〇8〇的解析度而 此時的YUV域影像信號由於需作移動向量的調整，其所 需滿足的像素數目還會大於192〇xl〇8〇，為了滿足不同解 析度的需求，影像信號處理器510將γυν域影像信號輸 入縮放器520，縮放器520會利用諸如内插法等將影像解 析度調整至適當大小。處理好的影像圖框信號不會直接輸 出，而會先儲存在輸出緩衝器540,輸出緩衝器54〇由輸 出緩衝控制器530控制，當緩衝控制器53〇控制輸出緩衝 器540輸出一影像圖框信號至輸出介面控制器55〇後，輸 出介面控制器550會根據該影像圖框的總和移動向量修正 該影像圖框信號以產生輸出影像。 請參考第7圖1 7圖為本發明另—實施例之流程 圖。本發明提出一種可防手震之影像處理方法，以下提出 一種較佳實施例，包含下列步驟： 步驟71〇:產生-貝爾圖樣影像，同時執行步驟72〇與步 =驟咖：取樣貝爾圖樣影像中—原色並利用Ga_或其 他轉換將貝爾圖樣影像中一原色轉換為原色資料. 步驟730:儲存複數個圖框之原色資料並比較前後二圖框 12 201027982 之原色資料以產生一移動向量表； —圖框之總和移動 步驟740:根據該移動向量表計算得出 向量； 步驟75Q:接收並暫存該總和移動向量，跳至步驟790; 步驟m爾圖樣影像轉換成m域之—影像信號； 步驟770 :將該影像信號轉換 / ^ 付俠取所幂之像素比例及畫面比 例並輸出一影像圖框信號；The image is converted into a YUV domain image signal that can be recognized by a general display device. Generally, the amount of data in the YUV domain image signal is much larger than that of the Bell pattern image, and the image signal processor 510 additionally performs some color, saturation, and Image effects such as contrast. Since the display device connected to the back end is different, the display resolution required to satisfy the image signal of the YUV domain is different. In the case of the Full HD standard, the resolution of 192〇χ1〇8〇 is required and the YUV domain at this time is satisfied. The image signal needs to be adjusted by the motion vector, and the number of pixels to be satisfied is still greater than 192 〇 xl 〇 8 〇. In order to meet the requirements of different resolutions, the image signal processor 510 inputs the γ υ ν domain image signal into the scaler 520. The scaler 520 adjusts the image resolution to an appropriate size using, for example, interpolation. The processed image frame signal is not directly output, but is first stored in the output buffer 540, and the output buffer 54 is controlled by the output buffer controller 530. When the buffer controller 53 controls the output buffer 540 to output an image map. After the frame signal is output to the output interface controller 55, the output interface controller 550 corrects the image frame signal according to the total motion vector of the image frame to generate an output image. Please refer to Fig. 7 to Fig. 17 for a flow chart of another embodiment of the present invention. The invention provides an image processing method capable of preventing hand shake. The following describes a preferred embodiment, which comprises the following steps: Step 71: Generate a Bell pattern image, and perform step 72 and step = step coffee: sampling the Bell pattern image. Medium-primary color and use Ga_ or other conversion to convert a primary color in the Bell pattern image into primary color data. Step 730: Store the primary color data of the plurality of frames and compare the primary color data of the previous frame 12 201027982 to generate a motion vector table. - the sum of the frames moves step 740: the vector is calculated according to the motion vector table; step 75Q: receiving and temporarily storing the summation motion vector, jumping to step 790; step m pattern image is converted into m domain - image signal Step 770: Convert the image signal to / ^ to obtain the pixel ratio and the picture ratio of the power and output an image frame signal;
步驟780 :暫存該影像圖框信號； 步驟790 .根據該總和移動向量修正該影像圖框信號。 准以上所述者，僅為本發明之較佳實施例而已，當不 能以此限定本發明實施之範圍’即依本發明申請專利範圍 及發明說明内容所作之等效變化與修飾，皆屬本發明專利 可能涵蓋之範圍。 【圖式簡單說明】 第1圖為本發明之防手震影像處理装置一實施例方塊 圖。 • 第2A圖為貝爾圖樣示意圖。 : 第2B圖為貝爾圖樣經取樣並轉換後之綠原色資料示 意圖。 第3圖為Gamma轉換時轉換前像素亮度值與轉換後 像素亮度值的函數曲線圖。 第4A圖為Gamma轉換前的像素綠原色亮度直方圖。 第4B圖為Gamma轉換後的像素綠原色亮度直方圖。 13 201027982 第5圖為第1圖中移動判斷單元的方塊圖。 第6圖為本發明實施例之詳細方塊圖。 第7圖為本發明另一實施例之流程圖。 ： 【主要元件符號說明】 ： 10影像處理裝置 20感測單元 30取樣轉換單元 ❿ 40移動判斷單元 50影像處理單元 4 1 0圖框緩衝器組 420移動判斷器 430移動向量產生器 4 3 1移動向量表 432影像穩定分析器 440移動向量緩衝器 5 1 0影像信號處理器 ^ 520縮放器 530輸出緩衝器 ： 540輸出緩衝控制器 550輸出介面控制器 14Step 780: Temporarily store the image frame signal; Step 790. Correct the image frame signal according to the summation motion vector. The above is only the preferred embodiment of the present invention, and the equivalent variations and modifications of the scope of the invention and the description of the invention are not intended to limit the scope of the invention. The scope of the invention patent may cover. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing an embodiment of an anti-shake image processing apparatus of the present invention. • Figure 2A is a diagram of the Bell pattern. : Figure 2B shows the green primary color data of the sampled and converted Bell pattern. Figure 3 is a graph showing the function of the pixel brightness value before conversion and the brightness value of the converted pixel during Gamma conversion. Figure 4A is a histogram of the pixel green primary luminance before the Gamma conversion. Figure 4B is a histogram of the green primary color of the pixel after Gamma conversion. 13 201027982 Fig. 5 is a block diagram of the movement judging unit in Fig. 1. Figure 6 is a detailed block diagram of an embodiment of the present invention. Figure 7 is a flow chart of another embodiment of the present invention. : [Main component symbol description] : 10 image processing device 20 sensing unit 30 sampling conversion unit ❿ 40 movement determination unit 50 image processing unit 4 1 0 frame buffer group 420 movement determiner 430 movement vector generator 4 3 1 movement Vector Table 432 Image Stabilization Analyzer 440 Move Vector Buffer 5 1 0 Image Signal Processor ^ 520 Scaler 530 Output Buffer: 540 Output Buffer Controller 550 Output Interface Controller 14