Waveform Coding

In waveform coding, we code the image intensity itself or some variation of image intensity such as the difference between two consecutive pixel intensities. Since the waveform itself is coded, the coders are simple both conceptually and computationally. Waveform coding techniques attempt to reconstruct the waveform faithfully without detailed exploitation of information specific to a particular class of signals.

PCM (Pulse Code Modulation)

Simplest waveform coding method in which intensity is quantized by a uniform quantiser: The basic PCM system codes not only intensity but also codes transform coefficients and image model parameters. Basic PCM System is typically used in obtaining an original digital image from an analog image in most digital image processing applications.

The spatial resolution of a digital image is primarily determined by the number of pixels (size). The size is chosen on the basis of resolution desired in an application.

For a digital image of Pixels has resolution comparable to 35 mm film

For a digital image of Pixels has resolution comparable to TV frame (2 fields)

For a digital image of and - used in video telephone.

As we decrease image size, resolution gets poorer and details are lost.

Bit rate used in original monochrome digital pictures is 8 bits/pixel. This preserves sufficient quality and intelligibility. Bits/pixel used as bit rate can be misleading. for e.g. if we obtain a digital image by sampling an analog image at a rate much higher than what can be resolved by HVS, the number bits/pixel can be reduced with out any visible resolution decrease by simply under sampling the digital image .Meaningful measure is bit/frame in single frame coding or bit/sec in case of image sequence coding.

If the input signal fluctuates sufficiently and if the quantization step size is small enough in the basic PCM system, the quantization noise can be modeled approximately as additive random noise.

For typical images, the quantization noise begins to be visible as random noise at about 5 to 6 bits/pixel. As we lower the bit rate below to 3 to 4 bits/pixel, the signal dependent quantization noise becomes visible as false contours due to luminance jump in the reconstructed image in regions where original image intensity varies slowly. This is shown in Fig (13 ). Several simple ways to improve performance of basic PCM system are discussed next.