International Journal of Engineering and Technology Innovation, vol. 11, no. 2, 2021, pp. 122-134 Image Compression Using Permanent Neural Networks for Predicting Compact Discrete Cosine Transform Coefficients Saleh Alshehri * Department of Computer Science and Engineering, Jubail University College, Saudi Arabia Received 26 December 2020; received in revised form 20 February 2021; accepted 22 February 2021 DOI: https://doi.org/10.46604/ijeti.2021.6925 Abstract This study proposes a new image compression technique that produces a high compression ratio yet consumes low execution times. Since many of the current image compression algorithms consume high execution times, this technique speeds up the execution time of image compression. The technique is based on permanent neural networks to predict the discrete cosine transform partial coefficients. This can eliminate the need to generate the discrete cosine transformation every time an image is compressed. A compression ratio of 94% is achieved while the average decompressed image peak signal to noise ratio and structure similarity image measure are 22.25 and 0.65 respectively. The compression time can be neglected when compared to other reported techniques because the only needed process in the compression stage is to use the generated neural network model to predict the few discrete cosine transform coefficients. Keywords: image compression, discrete cosine transform, neural networks 1. Introduction Image compression is an important field, especially in social media applications. In these applications, there are many requirements. Since most social media applications run on smartphones and personal devices, media content size and device processing speed are the most important requirements. These two requirements relate to each other, where a change in one affects the other. For example, if the media content is large, the hardware processing time will increase as compared to small-sized media content. Therefore, media content size should be as small as possible so that it can be stored and transmitted with minimal storage and processing power requirements. Despite advances in information technology and, in particular, in the storage and processor industries, digital visual and audio systems are always speed- and storage-hungry. This is because there are continuous advances in smart phones and their applications, such that better image and video quality are improved alongside advancements in hardware designs. For example, in 2021, it is assumed that multimedia video content will need about 70% of communication bandwidth [1]. This is an important observation because most social media applications run on smartphones. While smartphones are always improving, they face processor speed limitations [2] due to the difficulty in processor cooling [2]. Mobile phone processors can be improved and made more powerful in terms of speed, but the heating problem can impose the usage of a thermal management system, where the processor speed has to be reduced [2]. Keeping in mind the consumer’s need for better smartphone cameras and audio hardware, it is necessary to search for alternative and supporting solutions. Current mobile phone cameras have high resolution capability and can capture images of up to, and larger than, 4000×3000 pixels in size [3]. A color image of such size requires about 36×106 bytes of unsigned integers to be stored in raw pixels. Storing many such images requires large storage. * Corresponding author. E-mail address: saaas101@gmail.com