Design for Embedded Image Processing on FPGAs, 2nd Edition

7Point Operations

Beginning with this chapter, the focus will be on how specific image processing operations may be implemented on an field‐programmable gate array (FPGA). Chapters 7–11 will describe preprocessing and other low‐level operations. Chapters 12–14 will discuss intermediate‐level operations.

7.1 Point Operations on a Single Image

The simplest class of image processing operations is that of point operations. They are so named because the output value for a pixel depends only on the corresponding pixel value of the input image:

(7.1)

where is some arbitrary function. Point operations may therefore be represented by a mapping, or transfer function, as shown in Figure 7.1.

The simplest hardware implementation is to use stream processing to pass each input pixel through a hardware block implementing the function, as illustrated in the right panel of Figure 7.1. Since each pixel is processed independently, point operations can also be easily parallelised by partitioning the image over multiple processors.

In spite of their simplicity, point operations have wide use in terms of contrast enhancement, segmentation, colour filtering, change detection, masking, and many other applications.

7.1.1 Contrast and Brightness Adjustment

One interpretation of the mapping function ...