Circuit arrangement for recognizing a human face

• US 5,570,434 A
• Filed: 02/25/1994
• Issued: 10/29/1996
• Est. Priority Date: 09/05/1990
• Status: Expired due to Term

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Badique, Eric
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): KELLEY, CHRISTOPHER S
• Application Number:
  US08/202,843
• Time in Patent Office: 977 Days
• Field of Search: 382/1, 382/2, 382/56, 348/14, 348/19, 348/78, 348/390, 348/396
• US Class Current: 382/279
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation
• CPC Class Codes: G06V 40/167 using comparisons between temporally consecutive images
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: H04N 19/20 using video object coding

Abstract

A circuit for recognizing a human face in a sequence of video is provided. The circuit includes a block structuring circuit in which the video is structured block-wise and wherein two consecutive frames are subtracted from one another. The circuit also includes a unit for block-wise post-processing the difference frames. The post-processing unit includes a comparator for determining the areas of the difference picture which may correspond to a moving object, a mapping unit for forming a status picture by mapping each block of the difference picture which corresponds to a moving object onto a binary "one", all other blocks being mapped to a binary "zero".

The status picture is transformed by a transforming unit into a unidimensional feature function. The unidimensional feature function is then fed to a convolution unit to determine whether the feature function could include a "head-shoulder" scene. If so a plurality of marking units, mark all points of the status picture located within a predetermined rectangle in which the "face" may be present. Each picture element within the predetermined area is then refined.

Claims

1. A circuit arrangement for recognizing a human face in a sequence of video pictures, each picture comprising a plurality of picture elements, wherein said sequence of video pictures is structured into blocks of said picture elements, respective blocks of two consecutive video pictures are subtracted from one another so as to form a difference picture comprising a plurality of difference blocks each comprising a plurality of picture elements each having a standard value, said standard values of said difference picture elements of each of said difference picture blocks are summed so as to form respective standard value sums, and said difference picture is mapped onto a status picture formed by status bits by comparing the standard value sum of each of said difference blocks to at least one threshold value and mapping those of said difference blocks having a standard value sum exceeding said at least one threshold value onto a status bit having a first binary value, and mapping those of said difference blocks having a standard value sum not exceeding said at least one threshold onto a status bit having a second binary value which is inverse to said first binary value, said circuit arrangement comprising:

a) transforming means for transforming a first area of said status picture comprising first status bits having said first binary value into a unidimensional feature function in such a manner that the total number of these status bits within a line or a column of the status picture is determined;

b) convolution means for standardizing said unidimensional feature function by convoluting said feature function with a second function so as to form a convolution integral and to supply a recognition signal when said current convolution integral corresponds to ideal values within predetermined tolerance limits; and

c) marking means for determining, when said recognition signal is present, the center point of each line of said first area of said status picture so as to determine a vertical center axis, and to mark a plurality of points of said status picture located within a predetermined face area, which is cut in half by said center axis.

2. A circuit arrangement as claimed in claim 1, wherein

first filter means are provided, to change the status picture such that a status bit having the first binary value is inverted when the number of adjacent status bits having the same binary value is below a predetermined threshold.

3. A circuit arrangement as claimed in claim 1, further comprising:

smoothing means for changing the status picture such that corresponding status bits of at least two consecutive pictures are combined into one new status bit by a logic OR- combining operation.

4. A circuit arrangement as claimed in claim 1, further comprising:

second filter means for changing the status bit.

5. A circuit arrangement as claimed in claim 1, further comprising:

median-filter means for smoothing a one-dimensional feature function by means of a median filtering.

6. A circuit arrangement as claimed in claim 1, further comprising:

means provided for dividing the areas of a video picture corresponding to face areas into a consecutively numbered number of regions wherein

6a) the picture elements of a region are all included in the same class of picture data, and which are provided for,

6b) determining the size and the centres of gravity of these regions.

7. A circuit arrangement as claimed in claim 6, further comprising:

means for;

determining the distance of a triplet from centres of gravity.

8. A circuit arrangement as claimed in claim 7, further comprising:

means for checking at least two consecutively determined optimum triplets to see whether they agree within predetermined tolerance limits, and

if there is agreement, for determining all those picture elements which according to reference data belong to a human face.