FLIC: A Fast & Lossless Image Compression
Implementation for Hyperspectral Images

--- Project for course EECS 454

   In this project, our aim is to devise and implement an algorithm to compress a sequence of multi-spectral images. The source images are taken in a time sequence by a multi-spectral sensor on an aircraft. The aircraft is flying back and forth over a region, from north to south, until the whole region is covered. The following are the specific requirements of our algorithm. A lossless compression algorithm is required. Because these images are for geographical use, information needs to be preserved as complete as possible. That is to say, the original image can be identically recovered from the compressed image. A real-time compression algorithm is required. In the problem we studied, the compression is going to be done in real time during the image capturing. That is to say, once the sensor has an image ready, it is compressed, and then saved in the repository. Since the aircraft carrying the sensor is moving very fast and the frequency of taking images is high, the compression is expected to be done as fast as possible. Otherwise, the compression will be of no practical use.

Introduction

      Image compression is the process of reducing the number of bits required to represent an image. Image compression has been an active research area for more than 30 years. Image compression methods can be divided into two categories, namely, lossless compression and lossy compression. In lossless compression, the original image can be reconstructed from a compressed image without any loss of information. On the other hand, lossy compression will eliminate some original image information permanently after compression. Many algorithms have been proposed for both lossless and lossy image compression, for example, PNG, JPEG and wavelet transformation.
       Historically, lossless image compression inherited the theoretical framework of text data compression. However, due to the spatial and temporal characteristics of image data, different context-based models and predictive models can be applied to achieve better compression performance. In addition, the multi-spectral images in our domain have some special properties. Existing text and image compression algorithms have their own properties, but may not work efficiently in our problem domain. When faced with a specific problem, we can often come up with a simpler more efficient solution. It is practical for us to propose a new algorithm for geographical multi-spectral image compression.

People

Hongbo Jiang
Dan Liu
Guangyu Chen
Meng Hu

Submission