A FPGA-based implementation of JPEG encoder

The research in the domain of image compression increased significantly where the requirements of transmission images have raised enormously. Image compression is very important in digital image processing. It plays a crucial role in efficient transmission and storage of images. The most widely used method of lossy compression is JPEG standard. In this paper, we will discuss the implementation of JPEG encoder for Field-Programmable Gate Array (FPGA). The target device is Virtex V ML507. The JPEG encoder was synthesized with EDK designs at the clock frequency of 125 MHz. The implementation starts with the standard JPEG algorithm that is analyzed to extract the interesting functions that can be implemented in an FPGA: quantization, Discrete Cosine Transform (D C T) and Huffman coding. Once identified, these functions are implemented in software. The design can compress from a BMP to a JPEG image with displaying the compressed one on screen.

I. Introduction

Due to the technological advances, we are driven to manipulate the more and more important data in the transmission and storage. Despite the increase in storage capacity and the speed of Internet, reducing the size of data remains the major concern in transmission field. In the case of images data, the slightest one contains more than 5 million pixels that, without compression, would stock on over than 15MB file. Therefore, a large storage capacity is required for this huge images data. That it will be very expensive in terms of cost. While the set of information is low, the transmission of information becomes much easier. After that, it is advantageous to reduce the number of data. Among the reduction methods, compression proceeding is used without modifying the feature of the image. The main purpose of image compression is to reduce the amount of data required for digital transmission and there storage capacity [1]. The most used technique to reduce images data is JPEG standard that is a lossy compression technique. This standard loses some part of data on DCT (discrete cosine transforms) step.