10-08-2012, 12:23 PM
JPEG IMAGE COMPRESSION ON THE TEXAS INSTRUMENT VIDEO PROCESSING BOARD TMS320DM6437
JPEG IMAGE COMPRESSION ON TI320DM6437.pdf (Size: 726.86 KB / Downloads: 99)
Abstract
Image compression is getting more and more important in the electronic world with increased amount of bandwidth and storage requirement with increase in the image and video usage over internet. Pioneering advances in the image compression algorithms are important. The project discusses various algorithms that are currently available in the commercial market for its advantages and disadvantages. One of them is Joint Photographic Experts Group (JPEG) image compression standard. The final goal of the project was to implement JPEG compression on the TI’s Video development Platform DM6437 DVDP. MATLAB simulation for reading an image in appropriate color space and apply DCT (Discrete Cosine Transform), Quantization, and RLE (Run length encoding) for image compression. For the hardware part of project, Image was captured through camera. DCT, Quantization algorithms was performed on it and tested on the TI’s board EVM320DM6437.
INTRODUCTION
Image compression is the upcoming field with the improvement in the image and video quality generation by generation. High quality digital photography, other images and video processing applications are coming into the market every day. Uncompressed multimedia such as images, video and audio data take considerable storage space and bandwidth when transferred over network. So, efficient storage and transfer of these images are of great importance. Despite rapid improvement in the storage density, processor speed and even network bandwidth demand for data storage capacity and transmission bandwidth continue to outstrip the capabilities of the currently available technologies. Therefore, compression of image, audio or video before storing or communication through network is very important for efficiently using the resources.
Purpose of the project
This project uncovers the fundamental idea of compressing the images for saving data storage and transmission bandwidth. First, it give a the survey of available compression algorithms with their advantages and disadvantages and then, the project is focusing on the JPEG (Joint Photographic Experts Group) image compression standards and algorithms. The main goal of the project is to perform different steps of JPEG algorithm on the Texas Instrument’s DSP video development platform DAVINCI EVM320DM6437.
JPEG compression
The name “JPEG” stands for Joint Photographic Expert Group, the name of the committee who created the JPEG standard. The JPEG standard defines the standard steps
to compress an image in to a stream of bytes and decompress it again to generate the original image back. JPEG can be adjusted for the compression ratio and the image quality as per the user requirement.
JPEG provides a general purpose standard to meet the needs of almost all still images used in the computer world. This algorithm is designed to specifically discard the information that human eye cannot see easily. This can be done because slight change in the color are not perceived well by the human eye, while slight changes in the intensity(light and dark) can be easily detected. JPEG compression standard was evolved for compressing the color or gray scale still images like photographs, graphics and video stills. Although JPEG is now used to provide motion video compression (Motion JPEG), the JPEG standards makes no special provision for such an application.
Co-efficient Quantization
The DCT output matrix takes more space to store than the original matrix of pixels. If the input to the DCT functions consists of eight bit pixels values, the values that come out can range from a low of -1024 to a high 1023, occupying eleven bits. To discard an appropriate amount of information, the encoder divides each DCT output value by a quantization value and rounds the result to an integer. A quantization value of sixteen produces a quantized DC co-efficient with seven bit precision. The larger the quantization value, the more data is lost. The actual DCT value is represented less accurately. Quantization is simply the process of reducing the number of bits needed to store an output value by reducing the number of bits to represent integer results.
Each of the 64 positions in the 8x8 matrix has its own quantization value, with the higher order terms being quantized more heavily than the low-order terms that is, the higher-order terms have larger quantization values. Furthermore, separate quantization tables are employed for luminance and chrominance data, with chrominance data being quantized more heavily than the luminance data.