ADAPTIVE SIGNIFICANT DCT COEFFICIENTS FOR TEXTURE-RICH GRAYSCALE IMAGE COMPRESSION Neelamma K. Patil 1 , 1 Dept. of Telecommunication Engineering, KLES College of Engineering & Technology, Belgaum, Karnataka, India 1 neelammakletc@gmail.com Fax: 91-831-2441644 Suresh F. Murgod 2 , 2 Dept. of Electronics & Communication Engineering, KLES College of Engineering & Technology, Belgaum, Karnataka, India 2 sureshmurgod@gmail.com Lokesh Boregowda 3 , 3 Fellow Engineering, Honeywell Technology Solutions Lab, Bangalore Karnataka, India 3 Lokesh.Boregowda@honeywell. com V.R.Udupi 4 4 Dept. of Electronics & Communication Engineering, KLS Gogte Institute of Technology, Belgaum, Karnataka, India 4 vishwa_u@yahoo.com Keywords: Image Compression, Texture Features, Discrete Cosine Transform (DCT), Inverse DCT (IDCT), adaptive significant DCT coefficients. Abstract: Images/Videos require large storage space and bandwidth. The discrete cosine transform (DCT) is widely used in image and video coding schemes. By applying the DCT to image blocks and quantizing the DCT coefficients, efficient image compression can be achieved as employed in standard JPEG coder. However, encoding of all DCT coefficients leads to excess storage space and bandwidth requirement during transmission. Because, the transmission of redundant information which has no purpose to serve but demanding for more bandwidth. Texture images need to be compressed without the loss of perceptual and texture information thereby achieving high compression with greater utilization of bandwidth is a challenging task till today. In this paper we analyze the problem from the compression and redundant information reduction perspective for texture images by proposing an adaptive image compression technique. Experimentation has been carried out on different image formats successfully. The storage space and bandwidth during transmission is efficiently utilized by encoding significant DCT coefficients and thereby preserving texture and perceptual information in the reconstructed image. 1 Introduction Data compression techniques in all data-management systems play a key role as a leveraging technology. It reduces the storage requirements, transmission time and bandwidth, which makes the data management more effective and efficient [1]–[3]. Images/Videos require large storage space and bandwidth. Because, the transmission of redundant information in an image/video which has no purpose to serve but demanding for more bandwidth and storage space. The DCT is capable of compacting image energy efficiently into a small number of DCT coefficients, thus suitable for image compression by reducing redundancy with greater extent. It has been widely adopted by various still and moving image coding standards, such as JPEG and MPEG [4]. Images and video are handled nowadays to a great extent in compressed formats based on block DCT transforms. This reduces demand for high storage and transmission bandwidth. However, lossy compression methods are highly optimized for generating description of images in which highly relevant perceptual information is preserved and all non-relevant information is eliminated and number of bits for perceptual description is minimized. Among the existing compression techniques, the transform technique is more effective and simple. The DCT plays an important role in this. The transformation from spatial domain to spectral domain is performed by DCT to achieve greater compression. Based on quantization strategy, DCT coefficients of high frequency (low amplitude) in the transformed domain are discarded and significant coefficients are preserved to increase the compression ratio without loss of visual information. An intelligent and adaptive way of selecting significant coefficients is a challenging task and is in high demand [5]-[8]. The process of encoding all DCT coefficients and transmitting them requires huge bandwidth and serve no purpose since most of the coefficients are zeros. By analyzing, we have proposed an intelligent and adaptive technique for selection of significant coefficients by eliminating redundancy. Texture features computed for each image will adaptively select the significant coefficients by reducing bandwidth and retaining texture and perceptual information in reconstructed image. The proposed method is very simple and straight forward. It does not involve complicated calculations for the implementation. 2 Texture Features Extraction and DCT 2.1 Texture Features Texture is one of the important characteristics of an image and is used in identifying objects or region of interest.