Miniaturized - Provided Pictures

Found Pictures

Taken from: http://bighugelabs.com/onblack.php?id=1112863024&size=large

Taken from: http://canemexmachina.deviantart.com/art/San-Francisco-Skyline-120610419

Personal Pictures

A picture taken at the Grand Buddha at Ling Shan, China. For both of these I used the second method I wrote, which instead of accepting a user inputted line, accepts a user inputted area. This was useful for things like the giant buddha, where I wanted only the statue in focus, not the stairs below it or the mountains to the side of it.

Both of the above images were taken in Shanghai, China in the Yuyuan Gardens.

For project 2 I implemented the basic algorithm. It takes two points selected by the user to create a dof region, either vertical or horizontal. Additionally, it saturates the colors of the created image. In order to create this effect I sent my function an image and the coordinates of the two user selected points. The points were used to create a mask of dof region, whose width I played around with depending on the image. This mask was then inverted so I could use it to apply the gaussian blur (created using fspecial) to all parts of the picture except for the dof region. For my second method, I instead accept a user inputted region to keep in focus. The rest of the methods functions similarly to the first method I wrote, although instead of expanding a rectangular region continuously, it first uses the user inputted mask and then after the first iteration uses an expanding rectangle-shaped mask.

I found the effectiveness of the miniaturizer varied drastically from photo to photo. For each photo I tweaked the width of the focused area, as well as the increment the mask shrank (to create a smoother blur gradient). Additionally, I messed around with the strength of the gaussian blur.