Image matching device and method for motion pictures

• US 7,773,674 B2
• Filed: 09/28/2007
• Issued: 08/10/2010
• Est. Priority Date: 08/02/2002
• Status: Active Grant

• Current Assignee: KDDI Corporation
• Original Assignee: KDDI Corporation
• Inventors: Sugimoto, Osamu, Wada, Masahiro, Kawada, Ryoichi, Koike, Atsushi
• Primary Examiner(s): VO, TUNG T
• Application Number:
  US11/905,281
• Publication Number:
  US 20080031339A1
• Time in Patent Office: 1,047 Days
• Field of Search: 375/240.16, 375/240.12, 375/240.15, 375/240.13, 348/699, 348/700, 348/413.01
• US Class Current: 375/240.16
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/10021 Stereoscopic video; Stereoscopic image sequence
• CPC Class Codes: G06T 7/223 using block-matching
• CPC Class Codes: H04N 5/145 Movement estimation for video coding H04N19/51
• CPC Class Codes: H04N 7/014 involving the use of motion vectors motion estimation and compensation in video coding H04N19/51

Abstract

A matching section performs motion estimation upon a predetermined processing unit of input video (p) by, e.g., an iterative gradient method. A motion vector (v) obtained by the iterative gradient method is obtained by the expression v=α·Δv+v₀ (wherein v₀ indicates an initial displacement motion vector and Δv indicates a differential vector). A characteristic amount extraction section extracts a characteristic amount from the distribution of motion vectors obtained by the motion estimation. A parameter determination section determines a conversion parameter α applied to the next processing unit by the characteristic amount. If the characteristic amount is equal to or larger than a predetermined threshold, the conversion parameter α is determined to be larger (e.g., α=1). If the characteristic amount is smaller than the threshold, the conversion parameter α is determined to be smaller (e.g., α=0.1).

Claims

1. An image matching method for performing an image matching by using an iterative gradient method which iteratively estimates at least one of motion or parallax of a video on a block-by-block basis based on an initial displacement vector, comprising:

calculating a motion vector for a block in a video frame using the iterative gradient method, the block having a first block size;

dividing the block into a plurality of small blocks, each of the small blocks having a block size smaller than the first block size;

determining an initial displacement vector for each of the small blocks with an initial displacement vector determination section, the initial displacement vector being determined by selecting an optimum motion vector from among motion vector candidates including motion vectors of blocks neighboring the block having the first block size and a motion vector of a block in a previous video frame corresponding to the block having the first block size; and

calculating respective motion vectors for each of the small blocks by performing the iterative gradient method, based on the initial displacement vectors, on each of the small blocks to calculate the motion or parallax for each of the small blocks.

2. The image matching method according to claim 1, wherein the motion vector candidates include at least one of the motion vectors of the blocks neighboring the block having the first block size, the motion vector of the block in the previous video frame corresponding to the block having the first block size and a motion vector obtained by calculating motion vectors for neighboring blocks of the block in the previous video frame.

3. An image matching device which performs an image matching by using an iterative gradient method for iteratively estimating at least one of motion and parallax of a video on a block-by-block basis based on an initial displacement vector, comprising:

a first iterative gradient method performing means for calculating a motion vector for a block in a current video frame using the iterative gradient method, the block having a first block size;

an initial displacement vector determination section for determining an initial displacement vector for a small block obtained by dividing the block having the first block size into a plurality of blocks having a block size smaller than the first block size, the initial displacement vector determination section determining the initial displacement vector for the small block by selecting an optimum motion vector from among motion vector candidates including motion vectors of blocks neighboring the block having the first block size and a motion vector of a block in a previous video frame corresponding to the block having the first block size; and

a second iterative gradient method performing means for calculating a motion vector of the small block by performing the iterative gradient method based on the initial displacement vector determined by the initial displacement vector determination section.

4. The image matching device according to claim 3, wherein the initial displacement vector determination section comprises:

a vector memory for storing at least one of motion vectors calculated by the first iterative gradient performing means for blocks in the current video frame and motion vectors calculated by the first iterative gradient method performing means for blocks in the previous video frame; and

an initial displacement vector selecting section for selecting the initial displacement vector of the small block from among motion vectors read out from the vector memory.

5. The image matching device according to claim 4, wherein the initial displacement vector determination section further comprises:

a calculation section for calculating a motion vector based on motion vectors read out from the vector memory, and

wherein the motion vector candidates include at least one of the motion vectors of the blocks neighboring the block having the first block size, the motion vector of the block in the previous video frame corresponding to the block having the first block size, and the motion vector calculated by the calculation section.