JP2645387B2 - Image processing method - Google Patents

Description

Description: TECHNICAL FIELD [0001] The present invention relates to a digital video processing method and a video processing system, and more particularly to a video processing system effective for real-time video processing display and real-time image analysis.

[Background of the invention and its problems]

When a screen larger than the display area is held and a part of the screen is scroll-displayed, a VRAM corresponding to the display area is provided for the frame memory holding the previous screen,
A process of gradually rewriting the contents of the VRAM is performed. In the case of simply changing the display area of the screen, such processing can provide a practically sufficient speed. However, in the case of performing a very high-speed scrolling process in some kinds of image processing, the speed is insufficient. Cannot be used for the calculation between images.

[Object of the invention]

SUMMARY OF THE INVENTION The present invention has been made to solve such a conventional problem, and has as its object to provide an image processing method capable of performing high-speed scrolling of the entire screen and calculation between images before and after scrolling.

[Summary of the Invention]

In the video processing system according to the present invention, a frame memory is connected to an input / output of a processing system, and a registration address in an output-side frame memory is shifted in a column direction and / or a row direction with respect to a read address of an input-side frame memory. A means for shifting is provided.

Such a video processing system can be applied to, for example, large-scale smoothing, and large-scale smoothing can be performed on an area of a power of 2 by repeating the inter-image calculation of images before and after scrolling.

(Example of the invention)

Next, an embodiment of a video processing system according to the present invention will be described with reference to the drawings.

In FIG. 1, the video processing system has a plurality of frame memories PM1 to PM5, and the output of each frame memory is input to a plurality of selectors SEL1 to SEL3 in parallel. An operation unit CUL is connected to the subsequent stage of each selector, and can perform an inter-image operation between outputs of the selectors.

Addresses for reading and writing are designated in each frame memory by the address counters AC1 to AC5. For example, a read address of the frame memory PM1 and a write address of the frame memory PM2 are shifted and set in the column direction and / or the row direction. If the image of the frame memory PM1 is transferred to the frame memory PM2, the image obtained by scrolling the image of the PM1 is written to the frame memory PM2. At this time, the pixel data of PM1 passes through the operation unit CUL without any operation. Next, the images in the frame memories PM1 and PM2 are sent from the selectors SEL1 and SEL2 to the operation unit CUL, where the inter-image operation is performed, and the result is recorded in the frame memory PM3 or the like. The image of this PM3 is shifted in the column direction and / or the row direction and is recorded on, for example, PM4, and an inter-image calculation is performed between PM3 and PM4. If the reading and writing of the frame memory are performed at the same time, the processing time can be greatly reduced as compared with the case where the memory is rewritten. If the processing in the operation unit CUL is pipelined, the inter-image operation can be performed at an extremely high speed.

In the case of performing large-scale smoothing of an area of 2 ^k × 2 ^k (k ≧ 1), assuming that an image already smoothed for an area of 2 ^k−1 × 2 ^k−1 exists, as shown in FIG. ^First, an average value of four pixels obtained by separating this image by 2 ^{k -1} in the column direction and the row direction is ^obtained (unweighted smoothing). That is, the following calculation is performed,
This is set as one of these four pixels, for example, the pixel value at the upper left.

P _ijm 2 ^k = ｛P _ijm 2 ^k-1 + P _{(i +} 2 ^k-1 _{) jm} 2 ^k-1 + P _{i (j +} 2 ^k-1 _{) m} 2 ^k-1 + P _{(i +} 2 ^k-1 _{) (j +} 2 ^k-1 _{) m} 2 _k-1 ｝ / 4 Equation (1) In order to perform this processing, the image shown in FIG. 2 is recorded in the frame memory PM1, and this is stored in the column direction by 2 ^k-1 The image shifted by the pixel is recorded in PM2, the average value of the pixels corresponding to PM1 and PM2 is obtained by the arithmetic unit CUL and recorded in PM3. The image generated in this way is the image shown in FIG. 3 (B). Next, an image obtained by shifting the image of PM3 by 2 ^k-1 pixels in the row direction (FIG. 3 (c)) is recorded in PM4, and the average of PM4 and PM3 is calculated and recorded in PM5. The image recorded on PM5 is a smoothed image of Expression (1). Here, in order to obtain a smoothed image of 2 ^k-1 × 2 ^k-1 , (k
-1) It is evident that it is necessary to perform scrolling once, and therefore, if the scrolling is performed k times in the row direction and the column direction, a smoothed image for an area of 2 ^k × 2 ^k is obtained.

If the scroll pixels are simultaneously recorded in two separate frame memories, one of them is read with a shift of 2 ^k-1 pixels, and the other is read as it is, and the two are inter-image calculated, the number of scrolls can be omitted once.

Note that in a video processing system having a k × k region convolution processing unit including a delay circuit, the (k−1) × (k−) stored in the convolution processing unit is used.
1) It suffices to perform smoothing on the pixels and perform a smoothing operation on the pixels separated by k pixels in the column direction and the row direction. Normally, convolution processing is performed on a 3 × 3 area, so large-scale smoothing can be performed on a 4 × 4 area without scroll processing, and scroll processing is extremely important for smoothing on a larger area. Becomes

〔The invention's effect〕

As described above, in the image processing method according to the present invention, the frame memory is connected to the input / output of the operation unit, and the registration address in the output side frame memory is changed in the column direction and / or row direction with respect to the read address in the input side frame memory. Since means for shifting in the direction is provided, and the calculation unit performs calculation between images before and after scrolling, large-scale smoothing can be performed at high speed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a video processing system according to the present invention, and FIG. 2 is a block diagram showing a 2 ^k × 2 ^k region from an image smoothed for a 2 ^k−1 × 2 ^k−1 region. FIG. 3A is a conceptual diagram showing a processing method when performing the smoothing processing of FIG.
FIG. 3B is a conceptual diagram showing the result of scrolling the image of FIG. 2B in the row direction by 2 ^k−1 pixels, and FIG. 3B shows the average value of the image of FIG. 2 and the image of FIG. FIG. 3C is a conceptual diagram showing an image, and FIG. 3C is a concept showing an image obtained by scrolling the image of FIG. 3B by 2K-1 pixels in the column direction. PM1 to PM5 frame memory, AC1 to AC5 address counter, SEL1 to SEL3 selector, CUL arithmetic unit.

Claims (2)

(57) [Claims] 1. A digital image having m × n (m, n ≧ 1)
And an image processing method for generating a large smoothed image in an area unit of the number of pixels, wherein two images can be input simultaneously,
An arithmetic unit that can calculate the average value of the pixel values of the corresponding pixels in these two images, a frame memory connected to the input and output of this arithmetic unit, and a registration address to the output frame memory in the input frame memory. Means for shifting the read address in the column direction and the row direction by using a means capable of shifting the read address by one pixel in the column direction or the row direction to generate an image composed of smooth pixel values by an inter-image operation. An image processing method comprising repeating this operation m times in the column direction and n times in the row direction. 2. An image processing method according to claim 1, wherein an image which is not shifted in each operation and an image which is shifted by one pixel in a column direction or a row direction are simultaneously generated.