Method for Correcting an Acquired Image

Abstract

A method of correcting an image obtained by an image acquisition device includes obtaining successive measurements, G.sub.n, of device movement during exposure of each row of an image. An integration range, idx, is selected in proportion to an exposure time, t.sub.e, for each row of the image. Accumulated measurements, C.sub.n, of device movement for each row of an image are averaged across the integration range to provide successive filtered measurements, G, of device movement during exposure of each row of an image. The image is corrected for device movement using the filtered measurements G.

Claims

1. A method of correcting an image obtained by an image acquisition device including: obtaining successive measurements, G.sub.n, of device movement during exposure of each row of an image; selecting an integration range, idx, in proportion to an exposure time, t.sub.e, for each row of the image, averaging accumulated measurements, C.sub.n, of device movement for each row of an image across said integration range to provide successive filtered measurements, G, of device movement during exposure of each row of an image, and correcting said image for device movement using said filtered measurements G.

2. A method according to claim 1 wherein said integration range, idx, is calculated dependent on a sampling frequency F for said measurements of device movement and said exposure time t.sub.e as follows: idx=F*t.sub.e

3. A method according to claim 1 wherein said averaging comprises calculating said filtered measurements G as follows: G=(C.sub.n−C.sub.n−idx)/idx where C.sub.n is an accumulated value corresponding to measurement G.sub.n and C.sub.n−idx is an accumulated value corresponding to measurement G.sub.n−idx at the start of said integration range.

4. A method according to claim 1 wherein said measurements of device movement comprise any of: gyroscope, accelerometer or magnetometer measurements.

5. A method according to claim 1 wherein said image comprises a still image or an image within a sequence of images.

6. A method according to claim 1 wherein said rows of said image are acquired during successive exposure times, t.sub.I, said measurements of device movement corresponding to respective exposure times.

7. A processing unit for an image acquisition device arranged to correct an image according to the method of claim 1.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0018] An embodiment of the invention will now be described, by way of example, with reference to the accompanying drawings, in which:

[0019] FIG. 1 is a block diagram of an image acquisition device according to an embodiment of the invention; and

[0020] FIG. 2(a) and FIG. 2(b) illustrate the effect of acquisition device movement during exposure of respective images from a rolling shutter exposure.

DESCRIPTION OF EMBODIMENTS OF THE INVENTION

[0021] In embodiments of the present invention, IMU sensor data is integrated over the exposure time of the lines of an acquired image so tending to average the record of device movement. As a result, correction to each line of an image is calculated for average camera orientation during the exposure time of the line. Importantly, the range of the integrator (or the corner frequency of the integrator) is a function of the exposure time of the image.

[0022] In one embodiment the IMU sensor data comprises gyroscope samples and the method operates as follows: [0023] 1. Successive gyroscope samples G.sub.n are added to a cumulative buffer, where each new element is a sum of the current value G.sub.n and all previous sample values added to the buffer. The cumulative value corresponding to sample G.sub.n is denoted as C.sub.n [0024] 2. An index to the cumulative buffer is calculated dependent on the gyroscope sampling frequency F and exposure time t.sub.e as follows: idx=F*t.sub.e [0025] 3. An average gyroscope rate is calculated: G=(C.sub.n−C.sub.n−idx)/idx [0026] 4. The average rate G is used as the numerical integration of camera orientation. It is these filtered values G which are used subsequently in place of the trajectory of corresponding original samples G.sub.n for correction of corresponding lines of an image in an otherwise conventional fashion. It will be seen that this mode of filtering does not require any additional corrections and will automatically adapt to changing exposure time.

[0027] It will be seen that for short exposure images, idx will be short and so the linear approximation G of device movement during exposure time will provide similar values to the original samples G and so will be appropriate for correcting such images. On the other hand, as exposure times increase, the averaging will have the effect of not overcorrecting an image subject to high frequency vibration, but can still provide useful correction for images subjected to human hand shake or tremor.

[0028] While the above example has been described in terms of gyroscope samples, the same technique can be applied to sample values from all IMU sensors (gyroscope, accelerometer, magnetometer) in implementations where full sensor fusion is required.

[0029] The above approach can be employed whether images are acquired using a rolling shutter technique or not. Where lines of an image are exposed successively, then successive corresponding IMU measurements filtered as above can be employed to correct for device movement during exposure of those lines; whereas for an image where all lines are exposed at the same time, then the same IMU measurements filtered as above are applied to all lines.

[0030] The filtered IMU sensor signals described above can be employed in electronic image stabilisation (EIS) schemes such as disclosed in co-filed application Ser. No. 15/048,149 in place of raw sensor signals conventionally employed to mitigate problems caused by high frequency vibration of the camera during image capture.