JPH05236449A - Moving image processor - Google Patents

Abstract

PURPOSE:To reduce time for work and hardware cost by executing the detection of a scene change point as image analysis and the preparation of index informa tion while utilizing one part of information in the process of image compression. CONSTITUTION:Moving image data are compressed into spatial information by an orthogonal transformation part 12, and one part of the transformed result is stored in a first storage part 17a as the feature amount of a compressed image. Difference between adjacent frames in the image feature amount stored in this first storage part 17a is calculated by an arithmetic part 18, and obtained differential data are stored in a second storage part 17b. Corresponding to the differential data stored in this second storage part 17b, the scene change point is detected by a detection part 19, and the detection of the scene change point as the image analysis and the preparation of index information is executed by utilizing one part of information obtained in the process of image compression.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving image processing apparatus for detecting a scene change point in moving image data of temporally consecutive frames.

[0002]

2. Description of the Related Art In recent years, with the accumulation of moving images using a computer, the development of display technology, and the realization of multimedia information, moving images have been frequently used.

An image analysis method such as indexing, which is required to grasp and edit the contents of such a moving image,
At present, most of the fast-forward playback, cue function, indexing, etc. are done manually.
Further, some methods of performing image analysis by hardware have been proposed, but all of these are for performing a dedicated circuit, and thus the overall circuit scale becomes large.

[0004]

Since moving images require a huge amount of information, storage techniques such as image compression will be indispensable in the future in order to sufficiently handle moving images on a computer. .. Further, work such as moving image editing requires an image analyzing work of grasping the contents in a short time and extracting a desired portion from a huge amount of moving image data.

However, when the image compression and the image analysis are individually performed, dedicated hardware is required individually, and there is a problem that the circuit scale becomes large and the cost becomes high.

The present invention has been made in view of the above circumstances, and an object of the present invention is to reduce the circuit scale by using at least a part of a circuit required for image compression and image analysis, thereby reducing the cost. An object of the present invention is to provide a moving image processing device capable of reducing the above.

[0007]

That is, according to the present invention, there is provided an orthogonal transform unit for compressing spatial information of moving image data by orthogonal transform, and an image obtained by compressing a part of the transform result obtained by the orthogonal transform unit. A first storage unit that stores the feature amount, a calculation unit that calculates a difference between adjacent frames of the image feature amount stored in the first storage unit, and a difference data obtained by the calculation unit are stored. A second storage unit and a detection unit for detecting a scene change point based on the difference data stored in the second storage unit are provided, and a part of information obtained on the assumption of image compression is used. Since it is possible to detect a scene change point and create index information as image analysis, it is possible to significantly reduce the time required for the above work and the cost of hardware and the like.

Further, according to the present invention, the detection section detects the scene change point by using the difference data for the frame stored in the second storage section and a predetermined number of frames located before and after the frame. Thus, it is possible to reduce erroneous detection of a scene change point that tends to occur particularly in a scene with a lot of movement.

Further, according to the present invention, the detecting unit stores the frame stored in the second storage unit and a frame of a certain number of widths located in front of the frame, and the frame and a frame of a certain number of widths located after the frame. The change point of the scene is detected by the two types of difference data, and it can accurately correspond to the change from a scene with a lot of movement to a scene with a little movement or a change from a scene with little movement to a scene with a lot of movement. By doing so, it is possible to reduce the erroneous detection of a scene change point which is likely to occur especially in a scene with a lot of movement.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows the circuit configuration thereof. Moving image data input in the form of RGB or YUI from a data input unit 11 is sent to a discrete cosine transform unit (hereinafter abbreviated as "DCT unit") 12. .. The DCT unit 12 compresses spatial information of the input same image data by a discrete cosine transform that performs orthogonal transform according to its frequency component, and outputs the transformed data to the quantizing unit 13 and the DC component calculating unit 16.

The quantizing unit 13 quantizes the converted data obtained by the DCT unit 12 with a prescribed number of bits and outputs the quantized data to the encoding unit 14. The encoding unit 14 variable-length-encodes the converted data quantized by the quantization unit 13, further compresses the data amount, and then outputs the data to the compression result output unit 15. Compression result output section
Reference numeral 15 outputs the encoded converted data from the encoding unit 14 to an external recording medium, for example, as a final compression result.

On the other hand, the DC component calculator 16 is a DCT unit.
The average luminance of one frame is calculated by calculating the average value for one frame by the direct current (DC) component value for each block obtained from 12, and is output to the storage unit 17 and the inter-frame difference unit 18.

The storage unit 17 has three areas, a calculation result storage unit 17a, a difference result storage unit 17b, and a detection result storage unit 17c. The calculation result from the DC component calculation unit 16 is stored in the calculation result storage unit.
It is stored in 17a.

The inter-frame difference section 18 is a DC component calculation section 16
The difference between the calculation result from the calculation result and the calculation result of the previous frame stored in the calculation result storage unit 17a is calculated, and the difference result is sent to the difference result storage unit 17b of the storage unit 17 and stored therein.
The difference result stored in the difference result storage unit 17b is the detection unit.
Read to 19.

The detection unit 19 detects a scene change point by adopting a statistical method for the difference results sequentially stored in the difference result storage unit 17b, and stores the frame number and the number of changes as the detection result. It is stored in the detection result storage unit 17c of 17.

Thus, the detection result output unit 20 reads out the frame number of the scene change point and the number of changes stored in the detection result storage unit 17c, and outputs the result of the image analysis, for example, by multi-display. The operation of the embodiment having the above configuration will be described below.

The time-continuous image data input by the data input unit 11 is subjected to discrete cosine transform for each frame in the DCT unit 12. This Discrete Cosine Transform is used for spatial information compression. By orthogonally transforming image data in the frequency domain, spatial correlation is removed, and energy is concentrated in some coefficients to significantly increase the amount of information. It is possible to reduce. Specifically, one frame of target image data is divided into a plurality of blocks of N pixels × N pixels, and two-dimensional discrete cosine transform based on the following equation (1) is performed for each block. That is,

[0019]

[Equation 1]

In the conversion coefficient F (u, v), F (0,0) is a direct current (DC) component coefficient corresponding to the average luminance value of the block, and the other is an alternating current (AC) component coefficient. Will correspond to.

In this way, the converted data whose information amount is significantly reduced is quantized by the quantizing unit 13 to a predetermined number of bits, and the encoding unit 14 performs hyper-biased encoding. Is output to. These DCT
Since the technique relating to the compression of the image data from the unit 12 to the quantizing unit 13 and the encoding unit 14 is already well known, further details will be omitted.

On the other hand, of the converted data obtained by the DCT unit 12,
As described above, DC corresponding to the average luminance value of the block
The DC component calculation unit 16 inputs the component coefficient, adds the DC component coefficients of all the blocks of the corresponding frame at once, and then divides by the number of blocks to calculate the average luminance value of the entire frame. The calculated average luminance value is sent to the inter-frame difference section 18 as it is and is temporarily stored in the calculation result storage section 17a of the storage section 17.

The inter-frame difference section 18 is a DC component calculation section 16
The difference between the average brightness value sent from the device and the average brightness value of one frame before read from the calculation result storage unit 17a is calculated. This inter-frame difference generally indicates a value that stands out at a scene change point, and the calculated inter-frame difference is stored in the difference result storage unit 17b for each frame number.

After the processing of the inter-frame difference section 18 for all the frames of the moving image data is completed, the difference result storage section
The difference results for all frames stored in 17b are detected by the detection unit.
It is read out to 19, where the scene change point is detected by a statistical method. Specifically, the detection unit 19 calculates the average value and the variance value of the difference result, and the following equation (2), that is, [threshold value] = [average value] + [variance value] × [block pixel constant] N] (2)

The threshold value is calculated in accordance with, and threshold processing is performed on all the difference results. A scene number, a frame number, a scene length (time), and the like are detected as index information, and the processing result is stored in the detection result storage unit 17c.

The detection result output unit 20 calls the index information from the detection result storage unit 17c in response to the user's request, and outputs the image of the scene change point and other information by, for example, multi-display.

In addition, since the contents of the moving image are various scenes such as a scene with a lot of motion and a scene with little motion, which are continuous in time series, the inter-frame difference section 18 for all the frames of the moving image data. Instead of storing the output of the above in the difference result storage unit 17b, the difference result for the total 61 frames including the previous frame and the next frame, for example, 30 frames is stored in the difference result storage unit 17b, and this is read out by the detection unit 19 and the The scene change point may be detected by a conventional method. Specifically, as shown in FIG.
Calculates the average value and the variance value of the difference results, and
The threshold value is calculated by the formula, and the threshold processing is performed on all the difference results.

As described above, since the threshold value processing is performed by limiting the frame width required for detection as shown in the figure, the threshold value is made to correspond to the change of the image, and erroneous detection which is likely to occur particularly in a scene with a lot of movement. Can be reduced.

In the above embodiment, the moving image data is discrete cosine transformed and the scene change point is obtained by the DC component thereof, but not limited to the discrete cosine transformation, other orthogonal transform methods may be used. , A of discrete cosine transform
It is also possible to utilize the C component. Further, the same result can be obtained by obtaining the average luminance value of the specific portion of the divided frame instead of obtaining the average luminance value of the entire frame. In addition, the present invention can be variously modified and implemented without departing from the scope of the invention.

[0030]

As described above in detail, according to the present invention, an orthogonal transform unit for compressing spatial information of moving image data by orthogonal transform, and an image obtained by compressing a part of the transform result obtained by the orthogonal transform unit are provided. A first storage unit that stores the feature amount, a calculation unit that calculates a difference between adjacent frames of the image feature amount stored in the first storage unit, and a difference data obtained by the calculation unit are stored. Since the second storage unit and the detection unit for detecting the scene change point based on the difference data stored in the second storage unit are provided, a part of the information obtained on the assumption of image compression is used. Since it is possible to detect a scene change point and create index information as image analysis, it is possible to significantly reduce the time required for the above work and the cost of hardware and the like.

Further, according to the present invention, the detection unit detects the scene change point by the difference data for the frame stored in the second storage unit and the frames of a certain number of widths located before and after the frame. In particular, it is possible to reduce erroneous detection of a scene change point that tends to occur especially in a scene with a lot of movement.

Further, according to the present invention, the detection unit stores the frame stored in the second storage unit and a frame of a fixed width located in front of the frame, and the frame and a frame of a fixed width located after the frame. Since the scene change point is detected by the two types of difference data, it can accurately respond to the change from a scene with a lot of movement to a scene with a little movement, or from a scene with little movement to a scene with a lot of movement. It is possible to reduce the erroneous detection of a scene change point that tends to occur especially in a scene with a lot of movement.

[Brief description of drawings]

FIG. 1 is a block diagram showing a circuit configuration according to an embodiment of the present invention.

FIG. 2 is a diagram exemplifying another threshold value calculation method by the detection unit of FIG. 1;

[Explanation of symbols]

11 ... Data input section, 12 ... Discrete cosine transform (DCT)
Unit, 13 ... Quantization unit, 14 ... Encoding unit, 14 ... Encoding unit, 16 ...
DC component calculation unit, 17 ... storage unit, 17a ... calculation result storage unit,
17b ... Difference result storage unit, 17c ... Detection result storage unit, 18 ... Interframe difference unit, 19 ... Detection unit, 20 ... Detection result output unit.

Claims (2)

[Claims] 1. A moving image processing apparatus for detecting a scene change point in moving image data of temporally continuous frames, and an orthogonal transform means for compressing spatial information of the moving image data by sequential orthogonal transform, and the orthogonal transform. First storage means for storing a part of the conversion result obtained by the means as the feature quantity of the compressed image, and the difference between the adjacent frame of the image feature quantity stored in the first storage means is calculated. The present invention is characterized by comprising a calculating means, a second storing means for storing difference data obtained by the calculating means, and a detecting means for detecting a scene change point from the difference data stored in the second storing means. Video processing device. 2. The detecting means detects a scene change point based on difference data for the frame stored in the second storage means and a predetermined number of frames located before and after the frame. 1. The moving image processing device according to 1.