The Research for Compression Algorithm of Aerial Imagery

In this study, the new method of the JPEG image compression algorithm with predictive coding algorithm combining was proposed, effectively eliminates redundant information of the sub-blocks and redundant information between the sub-blocks and sub-blocks. Achieved higher compression ratio compared to the JPEG compression algorithm and a good image quality.


INTRODUCTION
Aerial imagery is becoming more and more closely associated with people's lives, is already widely used 3D city modeling, terrain exploration.The camera is usually installed in the satellites or aircraft, if you want real-time access to pictures, you must use a wireless real-time transmission, wireless transmission bandwidth is limited and aerial imagery for the amount of data, so image compression algorithm of high compression ratio and smaller loss of image quality is important.
JPEG image compression algorithms with predictive coding algorithm combined.First, the image is divided into various sub-blocks.Secondly, using JPEG compression to base-block, generate a prediction block using a predictive coding on the basis of the baseblock, Last, performs JPEG compression on the difference between the two, so as to achieve greater compression ratio, the study also simulation test of the compression and decompression process to draw the final compressed image effect.Hu (2003) have a research of the still image coding method with international standards.Zhang (1999) make a study of the image processing and analysis.Taubman and Marcellin (2002) have a research of the image compression fundamentals, standards and practice.Wang and Li (2006) study the MATLAB 7.0 graphic image processing.Wn and Li (2009) have a research of the MATLAB JPEG image compression method research.Kong and Chen (2008) have a research of the media technology and application development.
In this study, the new method of the JPEG image compression algorithm with predictive coding algorithm combining was proposed effectively eliminates redundant information of the sub-blocks and redundant information between the sub-blocks and subblocks.Achieved higher compression ratio compared to the JPEG compression algorithm and a good image quality.

TO CODING THEORY
The basic principle of the predictive coding: The basic idea of the prediction compression to eliminate redundancy is that the prediction compression by only the new information is encoded in each pixel of the extracted.It is because there was correlation between the pixels, so it makes the prediction compression possible.Prediction coding system the basic structure was shown in Fig. 1.In this study sub-block as a unit, to eliminate the redundant information between the subblock by the prediction.

The basic principle of JPEG compression encoded:
The basic structure of the JPEG encoder shown in Fig. 2, the decoding process is the inverse process of the coding process.
JPEG coding using the Forward Discrete Cosine Transform (FDCT), which is calculated according to operator (1): where, f (i, j) represents the color component values for each point.
Then the FDCT transformed frequency coefficients are quantized, the purpose of quantization is to reduce the magnitude of the coefficient of not "0" and to increase the number of coefficients of the value "0".While the human eye to low frequency component of the image is more sensitive than the high-frequency component, so the quantization table of the values in the upper-left corner is smaller than the values in the lower-right corner.In order to increase the number of consecutive "0", the quantized coefficients are Z-shaped arrangement.Therefore, in the Run Length Encoding (RLE), there are a lot of continuous "0".DC coefficients are compressed by Differential Pulse Code Modulation (DPCM) technology; AC coefficients are compressed using the Run Length Encoding (RLE).Then further compression of the results after the DPCM and RLE using entropy encoding.Finally, the tag code and encoded image data comprising the bit data stream.

This study in the principle of compression algorithm:
The basic idea of this algorithm is the JPEG compression algorithm and the predictive coding algorithm is combined to achieve the purpose of high compression ratio.The coding algorithm steps: Step 2: The image is divided into 8*8 sub-blocks and three types of sub-block are defined: base-block (I), the Prediction block (P) and Bidirectional prediction block (B).Beginning of the line as defined in the first sub-block as a base-block, the base-block appears interval is N. N subblocks of between base-blocks are the prediction block or bidirectional prediction block.M bidirectional prediction blocks are between base-block with the prediction block or prediction block with prediction block.The subblocks arranged in the basic structure shown in Fig. 3.The first sub-blocks are transformed using FDCT and then the use Table 1 and 2   shaped arrangement or the entropy coding to obtain a final code: where, D I (i, j) : The frequency coefficient of the current sub-block in front of the baseblock or prediction block D (i, j) : The frequency coefficient for the current sub-block is quantized Step 5: Bidirectional prediction generation of the sub-blocks: Similarly the first sub-block is FDCT transformed and then the light signal and the color difference signals is respectively quantized using Table 1 and 2, a quantization finished after is calculated using Eq. ( 4).And the results then later work of the JPEG compression algorithm: Z-shaped arrangement, entropy coding, etc., to obtain a final code word.
1 1 ( , ) ( , ) ( , ) ( , ) 2 where, D I (i, j) k-1 : The front of the base block, or prediction block quantization after the frequency coefficients D I (i, j) k+1 : For the back of the base block, or prediction block quantization after the frequency coefficient D (i, j) : To be generated after a bidirectional prediction block is quantized frequency coefficients In the decoding process is the inverse process of encoding, decoding the first base block and then according to the base-block, the prediction block is decoded, then a bidirectional prediction block is decoded according to the base-block and the predicted block and finally the individual sub-blocks are combined into a decoded whole images.

THE EXPERIMENTAL RESULTS
Compression algorithm simulation test: All of the parameters in this study are: the interval between the base block N is 15, the interval between the predicted block M is 3. Image resolution is 3872*2592 and 24 bit Bmp image used to compression test, the original image is shown in Fig. 4a, after compression and the decoded image effect shown in Fig. 4b.
In this study the image is divided into three categories for the different image content: aerial image of a single background (class A), the general aviation image (class B) and the aerial imagery of the image content changes quickly (class C).Respectively have been tested, 20 images of each class are tested and then the results were averaged and the result of compressed as shown in Table 3.
New compression algorithm for a higher compression ratio than JPEG still images can be obtained from the results can be seen.But the compression ratio and the contents of the image related, the compression ratio will decline when rapid changes of the image content.

CONCLUSION
This study presents a JPEG compression algorithm with predictive coding algorithm combining algorithm and its application to the static image compression.The new method on the basis of ensuring the quality of the image, greatly improve the compression ratio of the image.Especially in a single background image more good compression.
Fig. 1: A configuration diagram of the predict coding system Fig. 4: The original image and the processed image effect

Table 1 :
The quantization table of the light signals

Table 2 :
The quantization table of the color difference signals • The RGB image is converted to a YUV image • The image is divided into sub-blocks • Base-block compressed using the JPEG algorithm • Generate prediction block using a predictive coding on the basis of the base-block • Generating bidirectional prediction block Step 1: Because the human eye is more sensitive to light signals than the color difference signals, in this study first conversion RGB images into a YUV images and then quantizes the light signals and the color difference signals using two different quantization tables.The formula of RGB image is converted to a YUV image is:

Step 3 :
Base-block (I) is compressed using standard JPEG compression algorithm.This study using different quantization table to quantize the frequency coefficients of the light signals or the color difference signals.The quantization table of the light signals is Table 1.The quantization table of the color difference signals is

Table 3 :
Test data results