Module 3: Video Sampling Lecture 17: Sampling of raster scan pattern: BT.601 format, Color video signal sampling formats

Transcription

1 The Lecture Contains: Sampling a Raster scan: BT 601 Format Revisited: Filtering Operation in Camera and display devices: Effect of Camera Apertures: file:///d /...e%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture17/17_1.htm[12/31/2015 1:05:48 PM]

2 Sampling a Raster scan: BT 601 Format Revisited: One way to obtain a full digital video signal is to sample a raster scan which is a 1-D signal consisting of successive horizontal scan lines in successive frames (or fields for an interlaced raster). Therefore, the sampling interval along the scan lines determines the horizontal sampling interval. Choice of this interval depends on several factors: 1. The resulting horizontal sampling spacing must match the vertical spacing between scan lines. This implies that the sampling frequencies in horizontal & vertical directions are same; 2. The resulting samples in 3-D space should follow a desired sampling lattice. For example, if the samples on each frame or field are to form a rectangular grid, the sampling interval should be an integer divisor of line interval. These criteria were used in designing BT-601 digital video format described earlier. By selecting sampling interval according to, the samples on successive horizontal lines are vertically aligned, with horizontal & vertical sampling spacing being approximately equal. file:///d /...e%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture17/17_2.htm[12/31/2015 1:05:48 PM]

3 We assumed so far that the underlying video raster has a single luminance component. For the case of color video raster with one luminance and 2 chrominance components, straight forward approaches is to use same sampling frequency for all components. This is referred to as a 4:4:4 BT-601 video format. However as HVS is more sensitive to luminance information than chrominance, it is wise to sample chrominance components at lower resolutions than luminance. Ideally, sampling frequencies in all three dimensions should be reduced equally. However, because raster is stack of horizontal scan lines, reducing the sampling frequency for raster only lowers the horizontal sampling rate; the vertical & temporal rates remain unaffected. For eg. If sampling rate for chrominance component is (sampling rate for luminance), the temporal & vertical sampling intervals are the same for chrominance & luminance component; but the horizontal interval for chrominance is twice that of luminance. This is referred to as the BT-601 4:2:2 format. Similarly BT-601 4:1:1 format is obtained by down sampling the chrominance component by a factor of 4 along the raster scan. These are shown in figure below. file:///d /...e%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture17/17_3.htm[12/31/2015 1:05:48 PM]

4 (Figure 5) file:///d /...e%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture17/17_4.htm[12/31/2015 1:05:48 PM]

5 Now given a 4:2:2 signal, we can further down sample the vertical dimension by a factor of 2. This will lead to BT-601 4:2:0 format. Note 4:2:0 format cannot be directly obtained from the raster signal by uniform sampling. The 4:1:1 & 4:2:0 formats have the same overall sampling rates but, the 4:2:0 format can represent the video signal more accurately. This is because it captures the horizontal & vertical details in chrominance components with the same resolution that is half that for luminance component. On the other hand, with the 4:1:1 format, the vertical detail is represented with same resolution as luminance component, whereas the horizontal is represented by a quarter of that resolution. The discussion above assumes that 3 color components in raster signal are separated. To sample a composite color video, one must separate the individual Color components and then perform sampling. file:///d /...e%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture17/17_5.htm[12/31/2015 1:05:49 PM]

6 Filtering Operation in Camera and display devices: So far we discussed the sampling theory for K-D signals and 3-D video signals. We have also derived the required pre-filters and reconstruction filters for few typical video formats. Practical cameras and display devices are also known to perform filtering operations in a crude way. We discuss this problem in this section. Effect of Camera Apertures: Consider a camera that samples a continuously varying scene at regular intervals. Let represent the sampling intervals along the horizontal, vertical and temporal directions respectively. The underlying sampling lattice used is a cubic lattice. The respective sampling frequencies are ; and. Ideally the pre-filter should be a low-pass filter with cut off frequencies at half the respective sampling frequencies. In practice however the actual pre-filters implemented in typical cameras are far from ideal filters. file:///d /...e%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture17/17_6.htm[12/31/2015 1:05:49 PM]