Data link technology for a portable unmanned aerial vehicle

In this report we review the current state of the art regarding three of the most prominent image compression techniques, namely discrete cosine, fractal and wavelet transform coding techniques. Our interest in these techniques arises out of a desire to achieve as much compression as possible in the real time transmission of image data from the sensor payload onboard a portable Unmanned Aerial Vehicle (p-UAV) to a ground control station similar in form to a laptop computer. In particular, we find that compression ratios of greater than 30:1 are required in order to receive VGA images of 640X400 resolution and 24 bit colour at a TV frame rate of 25 Hz. This would be suitable for Line Of Sight (LOS) naval surveillance operations. Even greater compression ratios are required to receive images of 320X200 resolution and 8 bit grey scale at a frame rate of 1 Hz which would allow for transmission along a High Frequency (HF) data link in land based reconnaissance operations. An HF data link removes the need for LOS transmission, although it may be beyond the present capabilities of all three compression techniques to produce images of acceptable quality. For LOS transmission in the Very and Ultra High Frequency bands image compression would still be required, particularly if transmission were to occur at the TV frame rate. Even for lower frame rates image compression is desirable either to reduce the demand on the limited power source of the p-UA V, extend the vehicle's range of operation or make transmission more jam resistant. Of the three lossy techniques, the most popular is the Discrete Cosine Transform (DCT) technique, which is based on discrete Fourier transform theory. Although still undergoing further development as described in this report, the DCT technique has already met the standard for image compression put forward by the Joint Photographic Experts Group GPEG). The DCT is limited in the amount of compression that can be achieved without serious degradation of the data resulting in block artefacts appearing on reconstructed images. Thus we consider the non-conventional techniques of fractal and wavelet transform coding, which promise even higher compression ratios than the DCT. Fractal transform coding relies on the fact that many real world objects possess local self-similarity and can be described in terms of fractal transformations. These can be transmitted along a communications channel using less bandwidth than the pixel data of the original digital image. Fractal images not only provide a resolution independent image of the original, but can also yield very high compression ratios. However, at present it is questionable whether fractal coding is feasible for real-time applications, this report covers recent research directed at this question. In particular, we discuss recent attempts aimed at reducing the time expended in searching the domain blocks for each range block of the original image during encoding. We also describe the Accurate Fractal Rendering Algorithm which enables the fast decoding of video streams. These developments offer real hope that a fractal encoding/ decoding system will be available for near real-time applications in the not too distant future. Wavelets can be viewed as bumps that can be squeezed or expanded by dilation and shifted by translation. An arbitrary function can be decomposed into a series of wavelets forming a complete orthonormal set, the underlying principle behind wavelet transform coding. Wavelet coding has attracted much interest over the past few years, Inainly because it can bring about a reduction in the block artefacts associated with the DCT. Thus, it promises better quality reconstructed images at higher compression ratios than the DCT. At present it is unable to match the real-time performance of the DCT and may never reach those of fractal transform coding. With further advances in microprocessor technology and in optimising the software approaches described in this report, there is more than a possibility that this technique can be applied to near real-time applications soon. The value to Defence of this work is a greater understanding of the current state of lossy image compression techniques for possible implementation in communication systems where large amounts of data are required to be transmitted over narrow bandwidths. This examines examines data link requirements for a portable unmanned aerial vehicle. Crucial to the operation of such a data link is the development of suitable computer algorithms that are capable of significantly compressing and recontructing image data in a timely manner for viewing at a remote station. As a consequence of the near real-time requirement, we investigate recent advances in lossy data compression techniques concentrating on transform coding techniques involving the discrete cosine transform, fractals and wavelets. At present the discrete cosine transform is available on a microprocessor chip and can offer acceptable reconstructed images close to real-time with compression ratios of up to 35:1, but techniques techniques promise even higher compression ratios and possibly a near real-time capability in the not too distant future. DSTO