Floating-point camera

1. A camera comprising an image sensor, a massively parallel processor, and memory addressable by said massively parallel processor,

said camera being adapted to convert one or several image frames delivered by said image sensor into a temporary image representation in said memory substantively at a time when said one or several image frames are taken,

said temporary image representation being encoded as floating-point numbers comprising a mantissa and an exponent for at least one image channel, and

said floating-point numbers being encoded with more bits per pixel and channel than a number of bits per pixel and channel corresponding to a useful bit depth of said image sensor,

said camera being further adapted to compute a rendered image by said massively parallel processor, and retrieving pixels from said temporary image representation and performing operations to adjust the appearance of said temporary image representation in a parallel computation substantively at a time when said one or several image frames are taken.

2. The camera of claim 1 where said camera is portable.

3. The camera of claim 1 where said massively parallel processor is a graphics processing unit.

4. The camera of claim 1 where said floating-point numbers are being represented with at least 32 bits per a floating-point number.

5. The camera of claim 1 where said one or several image frames, comprising a plurality of pixels, delivered by said image sensor are delivered to said massively parallel processor as raw images with one channel where each pixel implicitly carries color information from knowledge of an arrangement of different color sensitivities of said plurality of pixels on said image sensor.

6. A camera comprising an image sensor, a massively parallel processor, and memory addressable by said massively parallel processor,

said camera being adapted to convert one or several image frames delivered by said image sensor into a temporary image representation in said memory addressable by said massively parallel processor substantively at a time when said one or several image frames are taken,

said temporary image representation being encoded with at least 24 bits per pixel and channel for at least one image channel, and

said temporary image representation being encoded with more bits per pixel and channel than a number of bits per pixel and channel corresponding to a useful bit depth of said image sensor, said camera being further adapted to compute a rendered image by said massively parallel processor, and retrieving pixels from said temporary image representation and performing operations to adjust the appearance of said temporary image representation in a parallel computation substantively at a time when said one or several image frames are taken.

7. The camera of claim 6 where said camera is portable.

8. The camera of claim 6 where said massively parallel processor is a graphics processing unit.

9. A machine-implemented method, comprising automatically retrieving one or several image frames from an image sensor,

converting said one or several image frames delivered by said image sensor into a temporary image representation in electronic memory substantively at a time when said one or several image frames are taken,

said temporary image representation being encoded as floating-point numbers comprising a mantissa and an exponent for at least one image channel, and

said floating-point numbers being encoded with more bits per pixel and channel than a number of bits per pixel and channel corresponding to a useful bit depth of said image sensor, and

computing a rendered image by a massively parallel processor, and retrieving pixels from said temporary image representation and performing operations to adjust the appearance of said temporary image representation in a parallel computation substantively at a time when said one or several image frames are taken.

10. The method of claim 9 where said image sensor, said electronic memory, and said massively parallel processor are part of a portable apparatus.

11. The method of claim 9 where said massively parallel processor is a graphics processing unit.

12. The method of claim 9 where said floating-point numbers are being represented with at least 32 bits per a floating-point number.

13. The method of claim 9 where said one or several image frames, comprising a plurality of pixels, delivered by said image sensor are delivered to said massively parallel processor as raw images with one channel where each pixel implicitly carries color information from knowledge of an arrangement of different color sensitivities of said plurality of pixels on said image sensor.

14. A machine-implemented method, comprising automatically retrieving one or several image frames from an image sensor,

converting one or several image frames delivered by said image sensor into a temporary image representation in electronic memory substantively at a time when said one or several image frames are taken,

said temporary image representation being encoded with at least 13 bits per pixel and channel for at least one image channel, and

said temporary image representation being encoded with more bits per pixel and channel than a number of bits per pixel and channel corresponding to a useful bit depth of said image sensor, and

computing a rendered image by a massively parallel processor, and retrieving pixels from said temporary image representation and performing operations to adjust the appearance of said temporary image representation in a parallel computation substantively at a time when said one or several image frames are taken.

15. The method of claim 14 where said image sensor, said electronic memory, and said massively parallel processor are part of a portable apparatus.

16. The method of claim 14 where said massively parallel processor is a graphics processing unit.