Object recognition apparatus and object recognition method

• US 7,224,842 B2
• Filed: 06/02/2005
• Issued: 05/29/2007
• Est. Priority Date: 06/29/2001
• Status: Expired due to Fees

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Kazama, Hisashi, Chujoh, Takeshi, Maruyama, Masayuki
• Primary Examiner(s): Bella; Matthew C.
• Assistant Examiner(s): Bayat; Ali
• Application Number:
  US11/142,317
• Publication Number:
  US 20050207620A1
• Time in Patent Office: 726 Days
• Field of Search: 382103-104, 382/233, 382/236, 382245-253, 348/143, 348/149, 348/148, 348/154, 348156-160
• US Class Current: 382/233
• CPC Class Codes: G06T 7/20 Analysis of motion motion estimation for coding, decoding, compressing or decompressing digital video signals H04N19/43, H04N19/51

Abstract

An object recognition apparatus comprises a moving picture decoding section which decodes data obtained by encoding a moving picture. A background image producing section produces a background image based on encoding information and a reproduction image. An object recognizing section sets a predetermined region to detect a feature quantity of an object, and carries out matching between the reproduction image and the background image, thereby recognizing the object in the moving picture.

Claims

1. An object recognition method comprising:

decoding data obtained by encoding a moving picture;

producing a background image based on encoding information derived during decoding and producing a reproduction image based on data decoded in said decoding; and

setting a predetermined region to detect a feature quantity of an object and carrying out matching between said reproduction image and said background image, thereby recognizing the object in the moving picture.

2. An object recognition method according to claim 1, wherein the setting step comprises performing a non-background determination process wherein whether mode information from the moving picture decoding section represents encoding or not is determined for each macro-block, and when the mode information represents encoding, a small block of vertical M₁ pixels and horizontal M₂ pixels is set in a macro-block, and whether the macro-block is a non-background or not is determined, based on a detection error between the small block of the reproduction image and the small block of the background image.

3. An object recognition method according to claim 2, wherein the setting step comprises performing a non-background determination process wherein whether mode information from the moving picture decoding section represents encoding or not is determined for each macro-block, and when the mode information represents not encoding, the processing goes to a next macro-block processing.

4. An object recognition method according to claim 2, wherein the setting step comprises performing a noise suppression process wherein when all the adjacent 8 macro-blocks of a macro-block which has been determined as non-background are background macro-blocks, they are determined as noise, and eliminated from non-background macro-blocks.

5. An object recognition method according to claim 4, further comprising:

storing the background image, said setting step comprising performing a process of updating the background memory wherein a portion of the background memory which has been determined as a background macro-block is updated in accordance with the reproduction image.

6. An object recognition method according to claim 4, wherein the setting step comprises performing a non-background inclusion process wherein a rectangular object is detected such that a macro-block which has been determined as a non-background is included, and it is determined whether or not the macro-block is included according to restriction in size of the object.

7. An object recognition method comprising:

encoding a moving picture;

producing a background image based on information derived during encoding and a local reproduction image based on information encoded in the encoding; and

setting a predetermined region to detect a feature quantity of an object, and carrying out matching between said local reproduction image and said background image, thereby recognizing the object in the moving picture.

8. An object recognition method according to claim 7, wherein the setting step comprises performing a non-background determination process wherein whether mode information from the moving picture encoding section represents encoding or not is determined for each macro-block, and when the mode information represents encoding, a small block of vertical M₁ pixels and horizontal M₂ pixels is set in a macro-block, and whether the macro-block is a non-background or not is determined, based on a detection error between the small block of the local reproduction image and the small block of the background image.

9. An object recognition method according to claim 7, wherein the setting step comprises performing a non-background determination process wherein whether mode information from the moving picture encoding section represents encoding or not is determined for each macro-block, and when the mode information represents not encoding, the processing goes to a next macro-block processing.

10. An object recognition method according to claim 9, wherein the setting step comprises performing a noise suppression process wherein when all the adjacent 8 macro-blocks of a macro-block which has been determined as non-background are background macro-blocks, they are determined as noise, and eliminated from non-background macro-blocks.

11. An object recognition method according to claim 10, further comprising:

storing the background image, wherein the setting step comprises performing a process of updating the background memory wherein a portion of the background memory which has been determined as a background macro-block is updated in accordance with the local reproduction image.

12. An object recognition method according to claim 10, wherein the setting step comprises performing a non-background inclusion process wherein a rectangular object is detected such that a macro-block which has been determined as a non-background is included, and it is determined whether or not the macro-block is included according to restriction in size of the object.

13. An object recognition method comprising:

encoding an input moving picture;

producing a background image based on information derived during encoding and producing a moving picture image based on said input moving picture; and

setting a predetermined region to detect a feature quantity of an object, and carrying out matching between said input moving picture image and said background image, thereby recognizing the object in the moving picture.

14. An object recognition method according to claim 13, wherein the setting step comprises performing a non-background determination process wherein whether mode information from the moving picture encoding section represents encoding or not is determined for each macro-block, and when the mode information represents encoding, a small block of vertical M₁ pixels and horizontal M₂ pixels is set in a macro-block, and whether the macro-block is a non-background or not is determined, based on a detection error between the small block of the input moving picture and the small block of the background image.

15. An object recognition method according to claim 14, wherein the setting step comprises performing a noise suppression process wherein when all the adjacent 8 macro-blocks of a macro-block which has been determined as non-background are background macro-blocks, they are determined as noise, and eliminated from non-background macro-blocks.

16. An object recognition method according to claim 15, further comprising:

storing the background image, wherein the setting step comprises performing a process of updating the background memory wherein a portion of the background memory which has been determined as a background macro-block is updated in accordance with the input moving picture.

17. An object recognition method according to claim 15, wherein the setting step comprises performing a non-background inclusion process wherein a rectangular object is detected such that a macro-block which has been determined as a non-background is included, and it is determined whether or not the macro-block is included according to restriction in size of the object.

18. An object recognition method according to claim 13, wherein the setting step comprises performing a non-background determination process wherein whether mode information from the moving picture encoding section represents encoding or not is determined for each macro-block, and when the mode information represents not encoding, the processing goes to a next macro-block processing.