402
403
Quality of SAR Image Processed by Fixed Point
Digital Operation (Commission II)
M.Ono and H.Tanaka
Mitsubishi Electric Corporation
325 Kamimachiya Kamakura, Japan
1. INTRODUCTION
In the digital processing of satellite borne synthetic aperture
radar (SAR) data, a vast amount of data should be processed in
very high speed. It is very advantageous if these amount of data
is reduced in order to minimize data storage area in the proces-
sor and to achieve high speed fixed point operation utilizing
this redueed size word length.
This paper describes the result on image quality assessment in
the data amount reduction by data quantization.
2. METHOD OF IMAGE QUALITY EVALUATION
To reduce data amount in the SAR data processing, quantization
of intermadiate data file is very effective. In the quantization,
limit of bit number reduction should be assessed Dy the evalua-
tion of image quality. Since both the dynamic range of processed
image and the radiometric resolution degrade due to the quantiza-
tion, they are the good measure to evaluate the degradation.
Dynamic range of the image can be calculated by the ratio of the
level of the brightest point to the level of the darkest point in
the processed image data when the scene is ground including urban
area. Radiometric resolution can be measured by the quantity à
where
& = Std.dev
=10 log, (1+ mean )
from the mean and standard deviation of smooth area. This
definition of radiometric resolution is based on Brooks and
Miller (1979).
3, SAR DATA PROCESSING
SAR images to be evaluated have been generated from the Seasat-1
SAR raw data which is Dbbit digital raw data.
To provide good understanding of evaluated image, the processing
block sequence depicting each processing stage is shown in Fiz.1.
The software has been developed by ourselves and modified for
intermadiate quantization. Range compressed complex data has been
quantized as follows
Imax , (Ain>Imax/2)
Ic= IFIX(Ain/(C*Std)*Ima/2*1max/2), (-Imax/2«Ain«cImax/2)
O (Ain<-Imax/2)
2