Go Down

Topic: sample and encode a picture (Read 380 times) previous topic - next topic

jfp

I want to take a camera-say a cheap webcam- and do the following..
Take a still picture. Then pixelate it down to a low level of resolution.
Form an array of rows and columns.
Then encode each  pixel in a column as  follows;
Top pxl  of column, say 5 Khz  audio, bottom pxl of column  500Hz audio.
The brightness of the pxl  is encoded as amplitude of the audio.
Is this  sampling and processing feasible with Arduino  variants?

Is it feasible to hack into  such a camera (because USB is offered)?
Advice, please, will be listened to most carefully.

AWOL

For reference, a 32 x 32 pixel image, with eight bits per pixel occupies 1 kbyte.
Is that low enough resolution?
"Pete, it's a fool looks for logic in the chambers of the human heart." Ulysses Everett McGill.
Do not send technical questions via personal messaging - they will be ignored.

PeterH

On a PC, this would be trivially simple. Simply reading an image is beyond the capability of most Arduinos but you might be able to do something if you choose one of the Arduinos with additional RAM, if you are prepared to accept very low image resolution and update frequencies. What sort of frequency and resolution are you aiming for? Will it be a colour image? Where do the audio frequency domain values come from?
I only provide help via the forum - please do not contact me for private consultancy.

jfp

Thanks for prompt responses. I am experimenting with  The vOICe  which is a well known sensory substitution system for blind users.  The system is capable of great subtlety of interpretation..
I wish to take it on at the  minimum level of resolution and simplification, to see how much functionality might remain at the lower ends.
I would hope to see it employed as a wearable collision avoidance/reaching/aiming device adjunct to the familiar white stick.
I am aware that Raspberry-pi  developers are applying themselves   ,via Android 4.0, to creating a truly wearable and low-cost ,full-performance version.
This being so, I still need to find out how to sample a B/W camera output , pixelate it at reduced resolution, encode the audio, and read out the  mixed audio chord as a left-to right scan of the picture columns.
The whole process is repeated at one frame per second.
I would appreciate  some  ballpark figures for the  possible  bandwidths, storage  and speeds required for a device as described. Is it really beyond the power of one of the  bigger `duinos?  Thanks...
(There is an excellent information  website at  seeingwithsound.com ,if you are interested.)


Grumpy_Mike

#4
Nov 10, 2012, 09:48 pm Last Edit: Nov 10, 2012, 09:54 pm by Grumpy_Mike Reason: 1
Quote
Is it really beyond the power of one of the  bigger `duinos?

Yes.
Unless you consider the new Due.

Have a read of this:-
www.robozes.com/inaki/dproject/report.pdf

Go Up