simon111's profile - activity

try to
IplImage DstImage0 = cvCreateImage(cvSize(rows, cols), IPL_DEPTH_8U, 1); ... cvTranspose(grayImage, DstImage0);
cvFlip(DstImage0,DstImage,flip_horizontal_or_vertical);

and you should modifiy code, so if you detected face, output some infos to logcat. I think you draw the face rectangle in a internal mat, so you can't see the rectangle in the final output view.

Actually, you need rotate your input video by 90 degree , flip is not enough,

As you noticed, many camera apps in Android devices only support landscape mode, fairly support protraint mode.

First, you can make your activity not to be a fullscreen app, I mean, remain the title bar or status bar. These UI elements will help you to know the real orientation of you activity.

Second, adjust your camera view suit for approperity activity oritentation. If you activity is portrait mode, you should do nothing. If you activity is landscape mode, you should rotate your input video by 90 degree ( transpose+flip ) , so you should detect the faces in portrait oritentaion.

It depends your project goal

You can

1) preprocess the 2 images to reduce the effect of the light conduction, rotation etc.

2) extract the feature for you image content, great feature is the most important part of image comparing.

3) compare the 2 feature sets, calculate the distance of the 2 feature vector. normalise the result

Hello experts:

I download the OpenCV 2.4.6, and compile it in Ubuntu12.04 with default compile option.

But I can't find the opencv_stitching program. Could you tell how to compile the opencv_stitching program ? Which compile option I should add ?

Thank you in advance

please try the following Python code:

import cv2
im=cv2.imread('image file name') # please put the actual image file name in the quothes
cv2.imshow('image', im)
cv2.waitKey(0)

could you provide some sample images?

Hi Experts:

I found initWideAngleProjMap in OpenCV source code, but I can't any documents on it.

Is this function predicted? Can I use this function with the newest version of OpenCV? Currently, I'm doing a project involved fish eye camera calibration. I used initUndistortRectifyMap function. It works not bad.

And i want to know is initWideAngleProjMap more suit for fish eye camera?

What are you want to do?

Are your want to process the screen image on an Android device? if so, why not just get a image of screen snapshot, and pass this image to the process program.

I recommend you include these dlls into your program, and you'd better load them dynamically.

If user system has an NVIDIA card, your program loads them.

calibrating and undistoring with opencv in c This is a great tutorial.

Use chessboard to get the camera initric matrix, and correct the image using the result matrix

Thank you, I will try

I'm calibrating a camera of bird's eye view, as you know, the distortion is very heavy.

After I read this tutorial

calibrating undistorting with opencv in c

I try to use a chess board to some experiences.

could you tell the whether chess board method is suit for a camera of bird's eye view?

think you in advance

Tranning is not very diffcult, but it spend long time (at least one week, depend on which computer you use), Opencv provides state-of-art tools for trainning. Check this link: http://docs.opencv.org/doc/user_guide/ug_traincascade.html

Samples is very important, you should prepare samples as many as possible. Maybe you need write a simple annotation tool to create samples.