Yi camera lens correction test

I was interested to find out what the difference was between filming with and without the Yi’s inbuilt lens distortion correction, once additional correction of the barrelling from the wide-angle lens had been corrected in After Effects.

CameraLensComparisons

The images on the left are taken without the Yi’s distortion correction, and all correction in the lower image in AE, using Optics Compensation effect, FOV 74, Reverse Lens Direction on. The images on the right are taken with the Yi’s distortion correction. Further, correction in AE, using Optics Compensation effect, FOV 48, Reverse Lens Direction on.

The result (once all barrelling is removed) is very similar between the two. However, as the correction entirely in AE means that the image is being digitally blown up, a better image quality is probably obtained by using the camera’s inbuilt correction.

Additionally, when using the camera’s distortion correction the raw footage has less distortion in more distant objects. For some footage without near objects with identifying barreling features (such as straight lines), it may be possible to avoid using the AE effect entirely.

Cone shadows continued

In viewing the cone shadow footage I realised that I need to decide if the projection should just be on a single surface, such as the back wall, or floor. This could use a shadow puppet arrangement, with the camera filming from behind a translucent screen. This could also work for filming the floor – the camera would be under the floor, which would also show the footprints of the forms, as well as their shadows. It would need to be strong enough to support the models, i.e. perspex.

The clip above takes the previous footage and only shows the back wall. I think this works well, although the shadow of the side wall and ceiling should be avoided by removal of these elements from the model. Again, the texture of the back wall is particularly strong, and a neutral surface may work better and act as a “screen” to receive the projection of the moving shadows.

Cone shadows

This is a test of a cone shadows model, using the found card as the “space” into which the cones’ shadows are projected. The intention is that the form of the cones is communicated entirely by the shadows, and this form is reconstructed by the mind of the viewer/reader.

The two edits below are made from the same piece of time-lapse footage (shot at 1frame/0.5 seconds), taken while the sunshine was very intermittent. The left hand edit just speeds up the entire recording to 120x real-time, while the right hand version edits out all of the sections where the sunlight has disappeared, and then speeds it up to 60x real-time, to allow more time for the moving shadows of the cones.

Continue reading

New found card model

Initially to test whether the Yi 1 will record for long continuous periods, I assembled a new model with the found card. As the Yi no longer has the back case it will not stand by itself, so I hastily assembled a lego stand. The camera successfully recorded continuously for several days and was manually turned off.

The triptych below shows the three days of recording, all with speed increased to play within 3 minutes. The majority of the “night” time (black screen) has been removed. The camera was set to lens distortion correction (as part of the lego stand was visible in full wide-angle), and did not need any further correction in after effects.

The oblique light across the undulating wall is particularly effective, and better than the last model where the light was direct.

Yi camera 1 working

The first Yi camera that completely stopped working in the last test is now working when both the back of the camera has been removed, including the battery. The picture above show both front and back removed, although I have since reattached the front (which allows me to use the makeshift focus ring), and it also works like this.

Yi camera problem investigations

I have now numbered each camera (just visible on the photo in the previous post), in the order they were purchased. Each camera has a different version of the firmware, and the later two are a different model to the first.

Camera 1 Wifi: YDXJ_2326644 SN: Z23A60252326644 FW: 1.2.13 Model: 23A
Camera 2 Wifi: YDXJ_3144435 SN: Z25L630ACN3144435 FW: 1.3.0 Model: 25L
Camera 3 Wifi: YDXJ_4476443 SN: Z25L640ACN4476443 FW: 1.4.8 Model: 25L

Using the found card model on the previous post, I noted each camera’s performance. Continue reading

Found card 3 Yi test model

y

In order to investigate the problems with some of the Yi cameras stopping recording, and to test the use of found materials for the models, I made a quick assembly from card from a recently purchased Ikea flat pack packaging. I will discuss the model and test filming here, and the investigation into the camera problems in the next post.

The material for the model was cellular card packing material, approx 30mm thick, small pieces of corrugated card (also from the Ikea packaging), and a piece of card from a box in which garlic bulbs had been delivered. I used a small sheet of perspex, borrowed from Alex from the front of his cabinet of curiosity (itself a form of diorama).

I used the camera mount wall from the three Yi model, so that I had ready made windows and camera locations. This meant that the width and height were the same as the three Yi model, although the depth was determined by the size of the pieces of card used for the floor surface (so slightly deeper than the previous model).

The assembly was covered in black felt to eliminate light ingress through the quickly constructed pinned joins.

I filmed with the camera’s inbuilt fish-eye correction disabled – this provided a wider field of view, but greater distortion of the image. As usual, I then “corrected” the footage via Adobe AfterEffects, using the “Optics Compensation” effect [reverse lens distortion box ticked, FOV: 77].

As previously, several of the cameras stopped working during filming (discussed further in next post). The footage below takes the complete recordings managed to be obtained from each camera, and alters their speeds so as to play within two minutes. Therefore the speed of each frame of the triptych is determined by the malfunctions of the cameras.

In the footage, the bright strip along the base of the back wall is caused by the thick perspex floor covering. I tried to cover all of the edges of the perspex to minimise this, but the effect is still there. Additionally, the use of perspex, rather than sticky film, as the shiny floor covering created very strong and sharp reflections of the sunlight. I prefer the mottled effect of the light reflected from the sticky film. I will either ensure I use this, or will source a “found” material that has a similar effect.