Rapid production of reduced-size images from compressed video streams
View Patent ↗Methods for fast generating spatially reduced-size images directly from compressed video streams supporting the coding of interlaced frames through transform coding and motion compensation. A sequence of reduced-size images are rapidly generated from compressed video streams by efficiently combining the steps of inverse transform, down-sampling and construction of each field image. The construction of reduced-size field images also allows the efficient motion compensation. The fast generation of reduced-size images is applicable to a variety of low-cost applications such as video browsing, video summary, fast thumbnail playback and video indexing.
1. A method for producing reduced-size images from a video stream, comprising:
selecting a reduced set of transform coefficients from each of a plurality of macroblocks by discarding higher-frequency coefficients in response to a reduction factor; and
performing an inverse transformation and field separation process on the selected coefficients to produce reduced-size top-field and bottom-field image blocks.
2. A method according to claim 1 , wherein:
the reduced set of transform coefficients comprises at least one transform coefficient from each of the plurality of macroblocks.
3. A method according to claim 1 , further comprising:
retrieving the reduced set of transform coefficients by performing the additional steps of:
decoding the macroblocks to produce the reduced set of transform coefficients in a quantized form; and
dequantizing the reduced set of transform coefficients.
4. A method according to claim 1 , wherein:
the reduction factor is expressed as a ratio M/N, where:
M and N are both positive, nonzero integers; and
M is less than N.
5. A method according to claim 4 , wherein:
each macroblock further comprises a plurality of transform blocks, each transform block comprising an N×N array of transform coefficients.
6. A method according to claim 5 , wherein:
the transform coefficients are Discrete Cosine Transform (DCT) coefficients.
7. A method according to claim 5 , wherein:
the macroblocks are field DCT encoded; and
the selected set of transform coefficients comprises 2M×M arrays of transform coefficients from each of a plurality of transform blocks.
8. A method according to claim 5 , wherein:
the macroblocks are frame DCT encoded; and
the selected set of transform coefficients comprise M×M arrays of transform coefficients from each of a plurality of transform blocks.
9. A method according to claim 1 , wherein:
the reduced-size image comprises a frame image.
10. A method according to claim 1 , wherein:
the reduced-size image comprises a field image.
11. The method of claim 1 , wherein each macroblock comprises at least two blocks.
12. A method according to claim 1 , wherein the macroblocks lie along a sampling path of a video frame, and the reduced size image is a line image, and further comprising:
producing a visual rhythm image of the video stream by stacking a plurality of line images alongside one another.
13. The method of claim 1 , utilized for at least one of video browsing, video summary, fast thumbnail playback and video indexing.
14. A device being arranged for executing a method as claimed in claim 1 .
15. A thumbnail generator comprising:
means for determining a reduction factor for a reduced size image;
means, responsive to said reduction factor, for selecting a reduced set of transform coefficients from each of a plurality of macroblocks by discarding higher-frequency coefficients; and
means for performing an inverse transformation and field separation process on the selected coefficients to produce reduced-size top-field and bottom-field image blocks.
16. A system according to claim 15 , further comprising:
means for adjusting the reduced-size image pixel values according to motion vector information within the video stream.
17. A system according to claim 15 , wherein:
the video stream is an MPEG-2 video stream.
18. A computer readable medium carrying a set of instructions which, when performed, execute a method comprising:
selecting a reduced set of transform coefficients from each of a plurality of macroblocks by discarding higher-frequency coefficient in response to a reduction factor; and
performing an inverse transformation and field separation process on the selected coefficients to produce reduced-size top-field and bottom-field image blocks.