I would recommend performing visualization intensive tasks and UIs on a separate front-end, given R and Matlab are not optimized to efficiently render charts and other visualizations.
If you are able to run WPF/Silverlight apps on your machine I can highly recommend SciChart (http://www.scichart.com/). It fulfills all your stated requirements. The library is targeting C#/WPF but I would imagine it to work well with Visual Basic or other .Net languages as well. Before proceeding I recommend to check out the Silverlight demo (no download or registration required) at SciChart Demo. I have extensively tested the library and now use it for a project and it fulfills all my requirements.
Regarding your requirements:
Zoom and scroll -> Is built in and works out of the box, you can either approach this issue through code-behind or set and customize properties and functionality within xaml (preferred).
Add vertical/horizontal/... lines -> can be easily done. I see annotation features added with each new release, however, the current library already offers extensive annotation capabilities. Take a look at the following (http://www.scichart.com/annotations-are-easy/, and http://www.scichart.com/questions/question/multiple-rollover-modifiers-mvvm)
Compute performances between points -> Not sure what you mean with that but if my assumption is correct then you can compute relative performance measures and generate a separate data series and render that because the axes are fully customizable. If you are referring to something different then please elaborate.
Add technical indicators -> You can add any data or time series that you desire. Technical indicators are not provided out-of-the-box but I recommend to always compute such metric on your own and simply use as data source to build a data series and render it on a chart.
Possibility to export the graph to the clipboard -> Not only that, but you can easily print or save charts in various formats as well. You can also easily store the specific template used for the chart as well as settings used so the next time you load the front-end your visualization settings will be restored.
I want to point out where this library shines in comparison to other libraries but also where I see room for improvement:
Maybe the strongest feature of SciChart is performance. Rendering millions of data points on a scatter plot, line chart, stock (OHLC) chart, and multiple other chart styles is not only high performant but also extremely simple to accomplish. I only know of one other library that plays in SciChart's performance league. I use this library to visualize tick-based bid/offer time-series data and I can render 5 million data points (I heard of guys who tried to render tens of millions of data points and accomplished such) and scroll, pan, and zoom at very satisfactory framerates. It lets me essentially zoom and magnify the time series to individual data points, the Date/Time axis can be customized to display millisecond or even higher resolution time stamps (something that could not be done with various other libraries when I tested them).
The library does not come free but the cost to acquire a license is reasonable in comparison with other libraries.
Through xaml and the ability to derive from certain base classes and very well designed bindability the styling and customization of annotations, axis types and labels, chart types, renderable series types, panes, synchronization of multiple charts, mouse/keyboard interactivity, among others can be accomplished without much effort.
I found the response time as well as quality of response of the support staff to be exquisite. I poked a lot of questions at staff and raised a number of support tickets at the beginning and I was generally attended to within 24 hours. That was even when I test trialed the product. Several questions only came up because I had very specific requirements about an issue regarding multiple chart panes.
While well tested, and apparently quite well received by developers at some financial institutions and apparently extensively used by scientists who require high performing visualization capabilities, the library is still a pretty recent venture. It may not offer the same quantity of content on its support forum as those who have been around for a decade or so longer, but again I found it pleasant to interact with support staff who was responsive and helped with certain details. So, immature is probably not the right term because it runs absolutely stable in production, but it is not the most mature library out there.
The library specializes in high performance charting for WPF/Silverlight and does its job excellently, however, it does not offer other controls such as diagrams, dashboard controls, list or textbox controls. I myself prefer to always work with the best that my budget can buy and use this library for charting, only. Other vendors shine in other categories and I can appreciate that.
In summary, I had very similar requirements as the ones you stated and this library did and still does a fantastic job at targeting my requirements. I use one chart implementation for high frequency data visualizations and another, which I styled very similarly to the Bloomberg chart library for overall low frequency price/volume time series visualization.
Disclosure: I at some point received non-monetary compensation in exchange for rendering a service to SciChart. But I believe this recommendation/review is biased only in that I truly believe in the quality of this library and the people supporting it because I myself extensively tested and now use it and found it to be meeting all my time series visualization requirements.
